UvA-DARE (Digital Academic Repository) Simulation of city evacuation … · 6th International Conference on Pedestrian and Evacuation Dynamics - PED 2012 Accompanying booklet abstracts

UvA-DARE is a service provided by the library of the University of Amsterdam (http://dare.uva.nl)

UvA-DARE (Digital Academic Repository)

Simulation of city evacuation coupled to flood dynamics

Mordvintsev, A.; Krzhizhanovskaya, V.; Lees, M.; Sloot, P.

Published in:PED 2012: accompanying booklet: abstracts of all presentations and posters: 6th International Conference onPedestrian and Evacuation Dynamics

Link to publication

Citation for published version (APA):Mordvintsev, A., Krzhizhanovskaya, V., Lees, M., & Sloot, P. (2012). Simulation of city evacuation coupled toflood dynamics. In PED 2012: accompanying booklet: abstracts of all presentations and posters: 6thInternational Conference on Pedestrian and Evacuation Dynamics (pp. [156-158]). Zürich: ETH [etc.].

General rightsIt is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s),other than for strictly personal, individual use, unless the work is under an open content license (like Creative Commons).

Disclaimer/Complaints regulationsIf you believe that digital publication of certain material infringes any of your rights or (privacy) interests, please let the Library know, statingyour reasons. In case of a legitimate complaint, the Library will make the material inaccessible and/or remove it from the website. Please Askthe Library: https://uba.uva.nl/en/contact, or a letter to: Library of the University of Amsterdam, Secretariat, Singel 425, 1012 WP Amsterdam,The Netherlands. You will be contacted as soon as possible.

Download date: 05 Jul 2020

https://dare.uva.nl/personal/pure/en/publications/simulation-of-city-evacuation-coupled-to-flood-dynamics(d70aa3a2-0d73-41ee-bd6b-d246007004d4).html

6th International Conference on Pedestrian and Evacuation Dynamics - PED 2012

Accompanying bookletabstracts of all presentations and posters

Opening

Wednesday, 6 June - 08.45

Ulrich Weidmann, Eidgenössische Technische Hochschule ETH, Zürich SWITZERLAND


Urs Walter, Fachstelle Fussgänger- und Radverkehr Stadt Zürich, Zürich SWITZERLAND

Keynote


Peter Jenkins, Building Design Partnership, Manchester UNITED KINGDOM Shaping the space: turning function into inspiration

It is impossible to extract from the station design process the importance of pedestrian flow management. Many aspects of the development of concepts influence (and are influenced by) the movement and activities of passengers:

• Makingspacefordiverseactivities• Relativeneedsandbehaviourofdifferentusers• Designingforspeedofmovement• Heritageconstraints• Workingwiththepedflowanalysis• Natural&guidedwayfinding• Thegatewayroleofastation• Thestationasadestinationinitsownright• Permeability• Safety&security

Thepresentationwillcovertheseandotherthemes,providinganexplanationofthe importanceofgaininganunderstandingofpedestrianbehaviourtoourwork.

Oral presentations


A.1 Tobias Kretz, PTV Planung Transport Verkehr AG, Karlsruhe GERMANY The effect of integrating travel time

A.2 Verena Reuter, Technische Universität Kaiserslautern, Kaiserslautern GERMANY On modeling groups in crowds: empirical evidence and simulation results including large groups

A.3 Gerta Köster, Technische Universität München, München GERMANY Validation of crowd models including social groups

B.1 Marco D‘Orazio, Università Politecnica delle Marche, Ancona ITALY Analysis of pre-movement times for the evacuation of university classrooms in the event of fire

B.2 Anne S. Dederichs, Technical University of Denmark, Lyngby DENMARK Simulex simulations on the evacuation of day-care centres for children 0-6 years

B.3 Edwin Galea, University of Greenwich, London UNITED KINGDOM Modelling evacuation using escalators: A London underground dataset

C.1 Cecile Appert-Rolland, University Paris-Sud, Paris FRANCE Experimental study of pedestrian dynamics

C.2 Laure Bourgois, IFSTTAR, Paris FRANCE Pedestrian agent based model suited to heterogeneous interactions overseen by perception C.3 Jan Dijkstra, Eindhoven University of Technology, Eindhoven NETHERLANDS Measuring individual’s egress preference in wayfinding through virtual navigation experiments

The effect of integrating travel time

Tobias Kretz, PTV Planung Transport Verkehr AG, Karlsruhe GERMANY

SHORTABSTRACTThiscontributiondemonstratesthepotentialgainforthequalityofresultsinasimulationofpedestrians when estimated remaining travel time is considered as a determining factor for the movementofsimulatedpedestrians.Thisisdonetwice:onceforaforce-basedmodelandonceforacellularautomata-basedmodel.Theresultsshowthatforthe(qualityof)simulationresultsthequestionifestimatedremainingtraveltimeisconsideredornotisofhigherrelevancethanthechoiceofthemodelingtechniqueforce-basedvs.cellularautomatawhichnormallyisconsideredtobethemost basic choice of modeling approach.

INTRODUCTIONIt has recently been argued that there is a common problem of pedestrian simulation software” which is thatmostifnotallavailabletoolsinthefieldoverestimatetheeffectivebottleneckasharpcorner(about90°andmore)imposesonalargegroupof(simulated)pedestrians(Rogsch&Klingsch,2010).Inotherwords: in simulations of pedestrian dynamics pedestrians cram too strongly when they move as a large grouparoundacorner.Thereforetheylosepaceandinconsequencetime.Theauthorsofsaidpaperarguebasedontheirexperienceandprofessionalbackgroundasplannersandapplicantsofsimulationsoftwarethat“Nocongestionshavebeenobservedinsuchcases(e.g.attheendofafootball-match)basedoncornermovement.”Indeedwhilethereappearstobenoscientificquantitativeinvestigationintothematteravailable,itisrathereasytofindvideofootageofvarioussuchsimulationsonthewebwhichsupportstheauthors’claim(Youtube,2011).

Theauthorsofsaidpapercorrectlyobservethattheproblemstemsfrompedestriansreceivingtheirmaindirectionofmovement(theirdesireddirection)fromafloorfieldorstaticpotentialwhichiscalculatedasdistancesfromaspottothedestination.Thereisnoreasontoassumethattheproblemisresolved or even just partially resolved if navigation is based on a navigation graph.

It appears that at least so far there is no model in which the interactions between pedestrians can fully compensate a desired direction which is computed in such a way. It is argued here without proof that the reason for this is that nearly all models of pedestrian dynamics are greedy algorithms (the property oflocaloptimalchoiceismoreobviousforcellthanforforcemodels,neverthelessanyshortestpathapproach is implicitly a greedy approach if in fact smallest travel time is intended) which either have acut-offdistanceofinfluencesor–ifinfluencesrangeinfinitelyfar–theinteractionbetweentwopedestrians is never related to their whole environment which consists of all the other pedestrians plustheirrelativepositionandmovementwithregardtothedestination.ClaimingthatthelatterisaproblemmeanstodemandthatsomehowSherif ’sobservationthat“thepropertiesofanypartare

determined by its membership in the total functional system” needs to be considered in models of pedestriandynamics,atleastforthesimulationofasubsetofsituations(Sherif,1936).

FROMSHORTESTTOQUICKESTPATHMOVEMENTPARADIGMRecentlytworelatedmethodshavebeenintroducedwhichachievewhathasbeenfoundtobedesirableintheprevioussection.Bothhaveincommonthatfirstforsmallareasatimedelayisestimatedif–hypothetically–apedestrianwouldwalkoverthatspot.Thetimedelayinbothmethodsiscalculatedfrom that this spot is currently occupied by a pedestrian. In one of the two methods the movement directionofthatpedestrianhasanimpactonthevalueofestimatedtimedelay.Second,these–small-area-time-delaysareintegrated(orbetter:summedup)startingatthedestinationandthereforecarryinginformationaboutjams(roughly)upstream.Theresultingfield(called“dynamicpotential”or“dynamicdistancepotentialfield”)isusedinthenextsimulationtimestep.

Theintegrationmethodaswellastheapplicationofthefieldinthesimulationisdifferentinbothapproaches.Thedetailsofthemethodsandespeciallytheirdifferencescannotbedescribedwithinthescopeofthisabstract.Weneedtorefertotheoriginalpublications:thefirstofthetwoapproaches(Kretz,2009)hasbeenconnectedtoacellularautomata-basedmodel(Kretz&Schreckenberg,2006),thesecondone(Kretz,Große,Hengst,Kautzsch,Pohlmann,&Vortisch,inprint)hasbeencombinedwiththeSocialForceModelaspublishedin(Johansson,Helbing,&Shukla,2007).Themessageinabottle for both approaches is that the main or desired direction is no longer calculated based on the shortestpath,butintothedirectionofestimatedleastremainingtraveltime,basedontheattemptof a realistic estimation of the options of an individual within an environment of static obstacles and moving–asindividualsorcrowds–pedestrians.

Makingpedestrianspreferablymovetimesteppertimestepintoadirectionofestimatedleastremainingtravelislabeledherethe“quickestpathparadigm”.

EFFECTSUsingbothmodelsasimplescenarioasgivenin(Rogsch&Klingsch,2010)issimulated;bothwithshortestpathaswellasquickestpathmovementparadigm.Furthermorethesameisdonewithanon-trivialgeometrywhichcouldoccurassuchina,howeversimple,real-worldapplication.Byvaryingtheparameters of the original dynamics model as well as the dynamic potential calculation it is shown that themajorimpactandimprovementcomesfromthedecisionifthequickestpathparadigmisappliedornot.Withregardtothedecade-longdiscussionontheprosandconsofthecellularautomatavs.theforceapproachonpedestriandynamicsthisisconsideredtobearemarkableresult.

REFERENCES

Johansson, A., Helbing, D., and Shukla, P. K. (2007). Specification of the Social Force Pedestrian Model by Evolutionary Adjustment to Video Tracking Data. Advances in Complex Systems . arxiv.org/abs/0810.4587

Kretz, T. (2009). Pedestrian Traffic: on the Quickest Path. Journal of Statistical Mechanics: Theory and Experiment . arxiv.org/abs/0901.0170

Kretz, T., and Schreckenberg, M. (2006). The F.A.S.T.-Model. ACRI 2006. Perpignan: Springer Berlin Heidelberg. arxiv.org/abs/0804.1893

Kretz, T., Große, A., Hengst, S., Kautzsch, L., Pohlmann, A., & Vortisch, P. (in print). Quickest Paths in Simulations of Pedestrians. Advances in Complex Systems . arxiv.org/abs/1107.2004

Rogsch, C., and Klingsch, W. (2010). Basics of Software-Tools for Pedestrian Movement—Identification and Results. Fire Technology .

Sherif, M. (1936). The psychology of social norms. Harper.

Youtube. (2011). Video Footage from various users: youtu.be/49HIZbFLPhg, youtu.be/jtKkHJXUVQY, youtu.be/LodYbDco0jY, and youtu.be/1WqnQjwAAac.James, John (1953): The Distribution of Freely-Forming Small Group Size. In: American Sociological Review, vol. 18, 569–570.

On modeling groups in crowds: empirical evidence and simulation results including large groups

Verena Reuter, Technische Universität Kaiserslautern, Kaiserslautern GERMANYBenjamin Bergner, Techn. Universität Kaiserslautern, Kaiserslautern GERMANYGerta Köster, Technische Universität München, München GERMANYMichael Seitz, Technische Universität München, München GERMANYFranz Treml, Technische Universität München, München GERMANYDirk Hartmann, Siemens AG, München GERMANY

Researchonpedestrianmovementstrivestomitigaterisksatlargeeventsorpublicinfrastructuresbybetterunderstandingtheflowofacrowd.Socialscientistsandmathematicalmodelersworktowardsthisgoalfromdifferentperspectivesrelyingonthetoolsoftheirtradesuchassurveysandobservational experiments on the one hand and mathematical formulations suitable for computer simulations on the other hand. In this paper both aspects will be combined to improve the modeling of groupsincrowdswithaspecialfocusonthekindofcrowdthatoneexpectstoattendnationalsoccerleague games. More precisely we are interested in regional evacuation around the soccer stadium in Kaiserslautern/Germany.

Inthecourseofourworkspecialattentionhasbeendrawntothepotentialimpactonsocialgroupswithincrowds.Ingeneral,investigatingandmodelinggroupshasbecomeanewfocusofcrowdresearchtriggeredbytheknowledgethatmassdynamicscannotsolelybeexplainedbythebehaviorofsinglemembersofthecrowd.Helbingetal.(2000)assume“amixtureofsocio-psychologicalandphysicalforcesinfluencingthebehaviorinacrowd”.Todaythereisevidence,bothempiricalandfromcomputersimulations,thatsocialgroupsinacrowdhaveaverysignificantimpactonpedestrianflowandthusoncriticalparameterssuchasevacuationtimes(Moussaidetal.,2010,Kösteretal.,2011).Thephysicalpartofthiseffectiseasilyexplainedbythefactthatpersonswhosticktogetherformbiggerparticles in a granular type of flow thus hindering the progress of evacuation.

Itisveryimportanttoknowthatthemembersofasocialgrouptrytostaytogetherevenin,orratherespeciallyin,potentiallydangeroussituations.Infact,makingcontactwithaffiliatedpersonstakesprecedenceoverindividualflight(Sime,1983).Thelossofcompanionsanduncertaintyabouttheirwell-beinghasadestabilizingeffectforpeopleaffectedbydisaster,whereasthepresenceofcompanionsisreducingfear(Mawson2005).Inourcasewearelookingat“psychologicalcrowds”likefansinasoccerstadiumwithastrongspiritoftogetherness(DruryandCocking,2007).Ifsocialgroupsaccidentallybreakupduringtheevacuationprocess,thesestrongtiesbetweenthegroupmembersmighthinderafastflow.Peoplemighteventurnagainstthestreamoffleeingpeopleinordertosearchfor lost companions.

Sofarmodelingresearchhasconcentratedonsmallgroupsof2to5personsneglectinglargergroups.Butsoccerfans,ourgroupofinterest,oftenvisitsoccergamesinagroupof5andmorepeople,e.g.friends,relatives,colleaguesandfanclubs.Thisposesthefollowingquestions:Dowealsoneedtomodellargegroupsincrowds?Dolargegroupsoccurfrequentlyenoughtohaveanimpact?Orisitperhapssufficienttoconsidersmallersubgroups?Whatmighttheimpactbeespeciallyinacrisissituation?Inthispaperwelookatthesequestionsfromasociologicalandamodelingperspective.

Atthesametimedistributionsofgroupsizeshaveyettobeinvestigated.UsingaPoissontypedistributionhasbeensuggested(ColemanandJames,1961;James,1953)andiscertainlyconvenientforimplementation,butisitrealistic?

ThispaperisajointeffortbysocialscientistsandmathematicalmodelerswithintheREPKAresearchproject(RegionalEvacuation:Planning,(K)Control,andAdaptation)toshedlightontheseareas.Empiricalsurveys(interviewsandobservations)havebeenconductedbysocialscientistsatlargeevents,andinparticularatasoccergame,togatherinformationontheoccurrenceandrelevanceoflargegroups.Thesurveysleadtofirstsuggestionsondistributionssuitableasinputformathematicalmodelers.Theempiricalresultssuggestthatgroupsdominatethecrowdatleastforthesportseventsweareinvestigating.Visitorscomewithfriendsandrelativesandonlyveryrarelyalone.Thereisevidencethatespeciallylargergroupsdooccurandshouldnotbeneglected.Wealsoascertaininourfirstobservationsthatlargergroupssplitupintosubgroups.Theymaintainaloosecoherence,whichhasanother impact than independent small groups.

Realizingthis,wecertainlyhavetobetterunderstandwhatimpactthecrowdcompositionhasincaseofanevacuation.Isthereapossibilitythatnewunexpectedbottlenecksemergebecausegroupmemberssticktogether?Isitpossiblethatknownbottlenecksaresoseverelyaggravatedthatescaperoutesareblocked?Inafirstattempttoanswerthesequestions,mathematicalmodelsexplicitlyrealizingdifferentcrowdcompositions(smallandlargegroups)areanappropriateapproach.Fourtypesofcrowdsarecompared:aggregatedcrowdsthatconsistentirelyofindividuals,crowdsthatincludesmallgroups,crowdsthatarecomposedoflargegroupsthatsplitupinsubgroupsandcrowdscomposedofindividuals,smallgroupsandlargegroupsaccordingtotheempiricaldistributionobtainedbythesurveys.Hereweconcentrateonasamplesituationwheretwopedestrianstreamscross.Ourresultsdemonstratethatnotonlygroupsingeneralbutalsolargergroupsinparticularaffecttheflowofacrowd.Thelargerthegroupsarethelesshomogeneoustheflowbecomes.

Wehopethatthisworkwillhelpandencouragefurtherinvestigationsconcerninggroupsincrowds.Anextstepmightbetolookatgroupswithspecialcharacteristicslikehandicappedpersonsorgroupswithspecialpurposessuchasrescueteamswalkingtowardsthesourceofthedangerinsteadoffleeingfrom it.

REFERENCES

Coleman, James S.; James, John (1961): The Equilibrium Size Distribution of Freely-Forming Groups. In: Sociometry, vol. 24, 36–45.

Drury, John; Cocking, Chris (2007): The mass psychology of disasters and emergency evacuations: A research report and implications for practice. University of Sussex, Department of Psychology. Brighton. Helbing, Dirk; Farkas, Illés J.; Vicsek, Tamás (2000): Simulating dynamical features of escape panic. In: Nature, vol. 407, 487–490.

James, John (1953): The Distribution of Freely-Forming Small Group Size. In: American Sociological Review, vol. 18, 569–570.

Köster, G., Seitz, M., Treml, F., Hartmann, D., Klein, W.: On the influence of group formation in a crowd, Journal of Contemporary Social Science, in print, to appear 2011/12

Mawson, Anthony R. (2005): Understanding Mass Panic and Other Collective Responses to Threat and Disaster. In: Psychiatry: Interpersonal and biological processes, vol. 68 (2), 95–113.

Moussaïd M, Perozo N, Garnier S, Helbing D, Theraulaz G (2010): The Walking Behaviour of Pedestrian Social Groups and Its Impact on Crowd Dynamics. PLoS ONE 5(4): e10047. doi:10.1371/journal.pone.0010047.

Sime, Jonathan D. (1983): Affiliative Behaviour during escape to building exits. In: Journal of Environmental Psychology, vol. 3 (1), 21–41.

Validation of crowd models including social groups

Gerta Köster, Technische Universität München, München GERMANYFranz Treml, Technische Universität München, München GERMANYMichael Seitz, Technische Universität München, München GERMANYWolfram Klein, Siemens AG, München GERMANY

Thedevelopmentofgroupmodelswithinmodelsofpedestrianmotionhasrecentlybecomeanewfocusofresearch.Thisinterestwastriggeredbyinsightfromthesocialsciences:Smallgroupsoftendominatethecrowdatlargeevents(Aveni,1977)andtheneedtoassociatewithfamilyandfriendsmaydominateoverflightinstincts(Sime,1983;Aguirreetal.,1998).Sofarmodelsofsmallsubgroupshavebeen suggested that well capture the movement towards a supposed common goal and ensure a certain spatialstructureofthegroupsforeasycommunication(Moussaïdetal.2010,Kösteretal.,2011).Ithasbeenshownbyboth,simulationsandmeasurementsthattheexistenceofsubgroupsinacrowdsignificantlyimpactsthebehaviorcrucialforthesafetyofthecrowdsuchastheflowatagivendensityoregresstimeswhenfacingabottleneck(Moussaïdetal.2010,Kösteretal.,2011).Itisthereforedesirablethatcrowdsimulatorsadoptthenewgroupmodelstobettermitigaterisksforexampleatlarge events or at public infrastructures.

However,tomakethisfeasiblereliablevalidationtestsmustbemadeavailable.Developersandusersalikeshouldbeabletocheckwhethertheadoptedmodelindeedcapturestheessentialcharacteristicsofacrowdcomposedofsubgroups.Asadesirablesideeffect,commonvalidationtestswouldmakesimulationtoolseasiertocompareandtheirrangeofapplicationeasiertoassess.Thiscanhelptoensureaminimumqualitystandardandthustofurthermitigaterisks.InthispaperwebuildonsomebasictestssuggestedbyKösteretal.(2011)forsmallsubgroupsandcomplementthembytestssuitablefor larger groups that split into subgroups.

Thefirststeptowardsourgoalistoagreeonwhatasocialgroupinacrowdis.Inthispaperwerestrictourselvestodescribegroupcharacteristicsthatimpacttheflowofacrowdandhencequantities,suchastheegresstimeatabottleneck,whichmaybecriticalforcrowdsafety(Kösteret.al.,2011).Wepresentthis very basic model of subgroups in a crowd for both small subgroups and larger groups in the form ofalistofrequirementssothatthechoiceofimplementationremainsfree.Hencethetestscanbeusedforsocialforcemodels,cellularautomataandothermodelchoicesalike.

Wethensuggestanumberofbasictests.Typicalquestionsare:Dogroupsstaytogether?Dogroupskeeptheirvelocitywhenwalkingalongafreepath?Dogroupmemberswalkabreastassuggestedbysocialcommunicationmodels?Aretheycapableofsplittingandreunitingwhilenavigatingaroundanobstaclelikeacolumn?Dolargergroupssplitupintosubgroupsassuggestedbysocialscientists(James,1951)?Dotheystillstaytogetherasawhole?Mostofourtestsarequalitativeintheirnatureand can be conducted by producing snapshots or short videos from the simulation output. In a number of cases it is also possible to introduce heuristic measurement parameters that allow running the tests inthemannerofautomaticJUnitteststhatareeitherpassedorfailed.Thisgreatlysimplifiesconsistent

development and testing.

Qualitativetestthatareusuallyvisualareveryimportantasastartingpoint:asimulationtoolthatfailstoensuregroupcohesion,forexample,cannotbeconsideredreliable.Howeveritisverydifficulttocomparedifferentmodelsandassesstheirrangeofapplicationonvisualobservationsonly.Thisiswhatmakesquantitativetestswhereanimportantoutputparameterismeasuredandcomparedagainstmeasurementsfromfieldorlaboratoryexperimentssovaluable.Foraggregatecrowdstheperhapsmostimportantquantitativetestisbasedonthedensity-velocityorequivalentlydensity-flowrelationshipinamovingcrowd.So-calledfundamentaldiagramsexpresshowthecrowdslowsdownwithincreasingdensity.Oberhagemann(2010)presentedgraphsoffundamentaldiagramsforsmallgroupsincrowdsextractedfromfieldstudies.Weproposetestsbasedonthesethefindings.However,sincetheresultsofthefieldstudyarenotyetgiveninanumericform,thesetestsarenotyettrulyquantitative.Theauthors therefore conducted a laboratory experiment where students are observed leaving a classroom asindividuals,inpairsandingroups(Kösteretal.2011). Wedemonstratethefeasibilityofallsuggestedtestswithourownsimulationmodelthatisbasedonacellular automaton.

REFERENCES

Aguirre, B. E., Wenger, D. & Vigo, G. (1998) A test of the emergent norm theory of collective behaviour, Sociological Forum, 13(2), 301–320.

Aveni, A. F. (1977) The not-so-lonely crowd: friendship groups in collective behavior, Sociometry, 40(1), 96–99.

James, J. (1951) A Preliminary Study of the Size Determinant in Small Group Interaction, American Sociological Review, 16(4), pp.474-477

Köster, G., Seitz, M., Treml, F., Hartmann, D., Klein, W., On the influence of group formation in a crowd, RSOC Journal of Contemporary Social Science, in print, to appear 2011/12

Köster, G., Seitz, M., Treml, F., Pfaffinger A., How to validate group models in pedestrian stream simulators: a proposal for basic tests. International Conference on Operations Research, Zürich,September 2011

Moussaïd M, Perozo N, Garnier S, Helbing D, Theraulaz G (2010): The Walking Behaviour of Pedestrian Social Groups and Its Impact on Crowd Dynamics. PLoS ONE 5(4): e10047. doi:10.1371/journal.pone.0010047.

Oberhagemann, Dirk. Risiko ”‘Veranstaltungen”’: Planung, Bewertung, Evakuierung, Rettungskonzepte, VdS Conference Proceedings, 2010.

Sime, J. D. (1983) Affiliative behaviour during escape to building exits, Journal of Environmental Psychology, 3(1), 21-41.

Analysis of pre-movement times for the evacuation of university classrooms in the event of fire

Marco D‘Orazio, Università Politecnica delle Marche, Ancona ITALYFrancesco Citro, Università Politecnica delle Marche, Ancona ITALY

Many studies show that people carry out activities not directly connected with the evacuation after hearingafirealarm:theytrytointerprettheinformationaboutthehazardbyinteractingwithotherindividuals;theywaittobereunitedwithotherpeople;theywastetimetryingtoretrieveobjects,andtheydonotgotowardsthenearestexitbuttotheonewheretheyenteredtheroom.Thistypeofbehaviormayslowdownorimpedetheevacuationofthebuildingtherebyincreasingtheriskoflossoflifeintheeventofafire.Insometypesofbuildings,thisriskisparticularlyhigh.Forexample,theproblemistypicalinschoolsanduniversities,wherethereisbothovercrowdingduringthehoursoflessonsandwidespreaduseofelectronicdevices(notebooks,tablets)aslearninginstruments.Moreover,inthesebuildingssomeoftheabovementionedtypesofbehavior(waiting,retrievingbelongings,etc.)mayoccurmorefrequently.

Severalstudieshavebeenconductedonsocialfactorsthatinfluencethe„evacuationtime“ofabuildingduringafire.AreviewofthesestudieswasconductedrecentlybyKobesetal.(2010).

D.Helbing,IFarkas(2009);D.Helbing,IFarkas(2000)describesuchbehaviorintermsofattractiveand repulsive social forces that lead individuals to achieve their goal (evacuation time) on the basis of interactionswithotherpeopleandtheenvironment.Althoughthisformulationisgenerallyrecognized,andusedforcommercialandresearchsoftware,Zhengetal.(2009),Koetal.(2007),showedthatbeforepeoplereallymovethereisaphasecalled“pre-movement”,definedasthetimeinwhichindividuals perform actions preliminary to moving: they try to interpret any information received aboutthedangersbyinteractingwithotherindividualspresent;theyarewaitforotherpeople;theytrytoretrievethingstowhichtheyareattached.Analysescarriedoutonpeople‘sbehaviorduringpreviousincidents,showthatthisphasedelaystheevacuationandthatithascausednumerousdeaths(PursereBensilum(2001),McConnelletal.(2010),KuligowskieMileti(2009)).PursereBensilum(2001)analyzeindetailthebehaviorofpeopleduringthisphaseinaseriesoffires.Onthebasisoftheiranalysis,theseauthorsshowhowthepre-movementprocesscanbedividedintotwostages:“recognition” (when people continue their activities) and “response” (when people carry out activities prior to evacuation).

Sincethe“pre-movementtime”stronglyinfluencestheoverallevacuationtime,severalexperimentshavebeenconductedwiththeaimtoquantifythisvalueinoffices,OveneCakici(2009),stores,ShieldseBoyce(2000),hotels,Kobesetal.(2010),cinemaandtheaters,NilssoneJohansson(2009),schools,Liuetal.(2009),Zhangetal.(2008).Onthebasisofexperimentaldata,PursereBensilum(2001)reporttheaverageresponsetimeforthe„recognition“and„response“phasesindifferenttypesofbuilding.Inparticular,theyshowanaveragetimerangingbetween0.03to1.05(decimalminutes)

forthe„recognitionphase“andbetween0.13and0.67(decimalminutes)forthe„responsephase“.Themaximum“pre-movement”timemeasuredinexperimentsis3.92decimalminutesforatheater(fromthe voice announcement).

However,thedatareportedbyPursereBensilum(2001)showawidedispersion,dependingonthetypeofbuildingandontheactivitiesinwhichthepeopleareengaged,LiueLo(2011).Infact,someauthorsagreethat,althoughthe“pre-movementtime”couldbeexemplifiedasanexplicitvalue,itisarandomvariablefollowingsomekindofprobabilitydistribution.PursereBensilum(2001)suggestaunimodaldistributioncharacterizedbyapositiveskewness.Normalandlognormaldistributionsarealso given.

Finally,studiesbyvariousauthorsagreeinhighlightinghowthe“pre-movement”timecanvarysignificantlydependingonthetypeofbuildingandtheactivitiesinwhichpeopleareengaged.Furthermore,theauthorsshowthatthe“pre-movement”timecouldbesignificantlyinfluencedbythe phenomena of “attachment to people” (waiting to be reunited with family members and friends) and“attachmenttothings”Zhengetal.(2009).Althoughthelatterisrecognizedtobeamajorcauseofdelayinpeople’smovementintheeventofevacuationthisaspecthasnotyetbeensufficientlyinvestigated,asindicatedbyKobesetal.(2010).Thereforeitwasconsiderednecessarytofurtherinvestigate this issue by conducting a simulated evacuation experiment in university classrooms. In theexperimentweevaluatedtheinfluenceof“attachmenttothings”,inparticulartoelectronicdevices(notebooks,tablets)thatarenowcommonlyusedintheseenvironmentsforeducationalactivities.Moreovercurrentteachingmethodsmakeincreasinguseofthesedevicesinclassroomsand,therefore,theevacuationofaschoolbuildingmaybestronglyaffectedbythisfactor.TheresultsarecomparedwithpreviousstudiesconductedbyZhangetal.(2008). InthisstudyweanalyzethebehavioroftwogroupsofstudentsinaFacultyofEngineering,followingtheactivationofafirealarm.Thestudyshowsthatinthesetypesofbuildings,duetothefactthatstudentsareinvolvedinactivitieswithelectronicdevices,“pre-movementtimes“areveryhighandare strongly influenced by users’ attachment to their belongings. In particular this study allows the distributionofpre-movementtimeandthespeedofpeopleduringtheevacuationtobecalculated.Thestudyconfirmedthebehaviorreportedintheliteraturecharacterizingtheinitialstagesofevacuationintheeventoffire.Inparticular,wedetectedtheexistenceofa„recognitionphase“(inwhichindividualscontinuewiththeiractivities)anda„responsephase“(inwhichindividualsperformpreparatorywork:turningoffelectricalequipment,etc.).

Inparticularweshowhowtheseactivities,atuniversities,becauseoftheintroductionofteachingmethodsthatusepersonalelectronicdevices,affectthe„pre-movement“timeandalsorepresentanobstacle for the rapid evacuation of students who did not wish to proceed before saving their data and tools.

Specifically,weconfirmthatthe„pre-movementtime“measuredcanbedescribedbyaWeibull-typedistributionwithapositiveskewness.

Simulex simulations on the evacuation of day-care centres for children 0-6 years

Anne S. Dederichs, Technical University of Denmark, Lyngby DENMARKLene Ulriksen, COWI A/S, Lyngby DENMARK

Performancebasedcodesaresuccessivelyintroducedacrosstheworld[Fleischmann,2011].Applyingthemethodsoffiresafetyengineeringrequiresmodelsdescribingtheevacuationofpeoplefrombuildings[Frantzich,1994].Datafromevacuationexperimentsareusedforvalidationandadjustmentsofsuchmodels[Kuligowskietal.,2010].Childrenaccountfor15-20%ofthewesternworld’spopulation[Larusdottiretal.,2010;Nordicstatisticalyearbook,2009]andevacuationmodelsshouldbevalidatedusingdatatakenfromthisgroupinordertoensurethesafetyofthispopulationandtofulfillthegrowingdemandon“equalegress”.OneofsuchmodelsfrequentlyappliedinfiresafetyengineeringisSimulex[Thompsonetal.,1994,1995].Themodeliswidelyappliedwithinfiresafetydesign.Themodelisdesignedalsotomodelevacuationofchildren.

ThepurposeofthecurrentstudyistocomparetheresultsfromsimulationswithSimulex1.2.[Frantzich,1998]onthemovementofchildrenintheage0-6yearsduringanevacuationwithrealdatafromexperiments[Larusdottiretal.,2010].Comparisonsoftotalevacuationtimesareshownanddiscussedinthisabstract,walkingspeedsonhorizontalplaneandonstairsarediscussedinthisabstract,theresultsofflowthroughdoorsandwalkingspeedsonhorizontalplaneandonstairswillbepresentedandanalyzedintheproposedpaper.

ComputersimulationsapplyingSimulex1.2of9differentday-carecentresarecarriedout.InDanishday-carecenterschildrenattheageof0-6yearsaredividedintotwogroups:youngerchildrenattheageof6monthsto2yearsandolderchildrenattheageof3to6years,i.e.crècheandkindergarten.Theresultsofthesimulationsarecomparedwithdatafromrealexperimentsfromevacuationexercises[Larusdottiretal.,2010]atthesameday-carecentres.

FollowinggeneralassumptionsaregiveninSimulex:• Eachpersonisassignedanormalunimpededwalkingspeed• Walkingspeedsarereducedaspeoplegetclosertogether• Eachpersonheadstowardsanexitbytakingadirectionwhichisatrightanglestothecontours shown on the chosen distance map• Overtaking,bodyrotation,sidewayssteppingandsmalldegreesofback-steppingareall accommodated.

Inthesimulationscarriedoutwithinthecurrentstudyoccupantsaredefinedaschildrenonly.Inthesimulations,allthechildrenarewalkingsinceitisnotpossibletospecificallydefinerunningchildren.Furthermore,Simulexdoesnotdifferentiatebetweendifferentagegroups;occupantsarejustdefinedaschildren.

Theday-carecentershavethreespiralstairs.Inthemodelthestairsaredefinedwiththetruelengthofthestaircase,theslopelengthaddedthelengthofthelandings.Thewidthofthespiralstairshouldbedefinedsonarrowthatitisnotpossibleforoccupantstoovertakeeachother.Stairsareshownasstraightstairs[Frantzich,1998].

Theegresstimeisdefinedasthetimefromthefirstoccupantstartsmoving,untilthefirstoccupantrespectivelythelastoccupantexitstheday-carecentre.Reactionanddecisiontimeareassumedtobeneglectableduetotheteachers‘knowledgeabouttheexercises.Thewarningtimesfoundintheexercisesisaddedtothewalkingtime.Figure1showsthetotalevacuationtimesfromthesimulationscompared to the times found in the evacuation exercises.

Theresultsshowthatthesimulatedtimesarelowerthanthetimesfromtherealdata.Thisappliesbothtothetimesforthefirstpersonoutandthelastpersonout.However,thereisabettercompliancebetweenthetimesforthefirstpersonoutwhichcouldalsobeexpected.Thesmallgapbetweensimulation and experiment can partly be explained by the fact that the reaction and decision time are neglected.ThedifferencebetweenthetimesforthefirstpersonoutintheexercisesrespectivelySimulexvariesbetween7secondsand56seconds.Thedifferenceon56secondsisforaday-carecentrewherethe children put on outerwear before leaving the building.

Resultsofthewalkingspeedinhorizontalplanewillbeshownintheproposedpaper.TheresultsfromtheexercisesgiveahighertravelspeedfortheolderchildrenrunningthantheaveragewalkingspeedfoundinSimulex.ThetravelspeedfortheyoungerchildrenrunningcomplieswellwiththeaveragewalkingspeedfoundinSimulex.However,onlyafewchildrenandnotenoughconsistencytousethenumbersine.g.adesignphaseforabuilding.Furthermore,thespreadoftheexperimentaldataofthetravelspeedsislargerthanfromthesimulations.Exceptforthefewchildrenrunning,thewalkingspeedfoundinthesimulationsinSimulexishigherthanthewalkingspeedfoundintheexercisesforbothagegroups.Thewalkingspeedsfromthesimulationscorrespondbetterwiththeresultsfromtheexercises for children running but still not completely.

Familiaritywiththeescape-routematteredwithrespecttotravelspeedalongthethreestairs.Itisnotpossibletosimulateoccupant‘sknowledgeandfamiliaritytoagivenstair.Exceptfortheinitialescapeinthefamiliarstaircase,thesimulationsoverestimatethetravelspeedonstairs.

REFERENCES

Frantzich, H., En modell för dimensionering av förbindelser för utrymning utifrån funktionsbaserede krav, Tech. rep., Department of Fire Safety Engineering, Lund University, (In Swedish) (1994).

Frantzich, H.: „Användarmanual till Simulex ver 1.2“, Inst. för brand-teknik, Lund University, Sweeden (1998)

Fleischmann, C.M., Is Prescription the Future of Performance Based Design?, IAFSS Symposium Maryland, 2011

Kuligowski E. and Peacock R., A review of building evacuation models, National Institute of Standards and Technology Technical Note 1471, 2010, 156pp.

Larusdottir, A.R. and Dederichs, A.S., (2010) Evacuation of Children: Movement on Stairs and on Horizontal Plane, Fire Technology special issue. http://dx.doi.org/10.1007/s10694-010-0177-6

Nordic statistical yearbook, Nordic Council of Ministers, Nordic Council, 2009.

Thompson, P. A. & Marchant, E. W. (1994). Simulex; Developing New Computer Modelling Techniques for Evaluation. In Fire Safety Science -- Proceedings of the 4th International Symposium (pp. 613-624).

Thompson, P. A. & Marchant, E. W. (1995). Testing and Application of the Computer Model ‚Simulex‘. Fire Safety Journal, 24, 149-166.

Modelling evacuation using escalators: A London underground dataset

Edwin Galea, University of Greenwich, London UNITED KINGDOMMichael Kinsey, University of Greenwich, London UNITED KINGDOMPeter, Lawrence, University of Greenwich, London UNITED KINGDOM

Thedevelopmentandexpansionofunderground(subway)stations,oftenlocateddeepunderground,hasbeenpossiblewiththeintroductionofescalatorscapableofefficientlytransportinglargevolumesofpeople[1].Asaresult,undergroundstationsarereliantuponescalatorsforcirculationandinmanycasesemergencyevacuation.Despitethis,fewstudieshaveattemptedtoquantifyhumanfactorsassociatedwithescalatorusage(microscopicanalysis),themajorityofpaststudiesfocusingonestablishingcapacity(macroscopicanalysis)ratherthanusagebehaviours[2-4].Assuch,itisuncertainhow human factors associated with escalator usage impact escalator performance in both circulation and evacuation situations. It is also uncertain whether human factors associated with escalator usagehasaculturalcomponent.Toaddresstheseissues,escalatorhumanfactorsdatawithinthreeundergroundstationsinSpain(Barcelona)[5],China(Shanghai)[6]andEngland(London)havebeencollected.Ineachlocationthesamemethodologyfordatacollectionandanalysiswasused.ThispaperpresentsanoverviewoftheanalysisfortheEnglishdataset.Furthermore,usingthedatacollectedalongwiththenewlydevelopedescalatormodelavailablewithinthebuildingEXODUSevacuationsoftware[6,7],aseriesofevacuationscenariosofahypotheticalundergroundstationarepresented.Theevacuationanalysisisintendedtoexploretheimpactofusingescalatorswithavarietyofrealistichuman factors.

Thehumanfactorsdatacollectedrelatestoescalator/stairusage,walker/riderusage,sideusage,walkerspeedsandboardingflow-rates.Intotal,datarelatingto11,019pedestrianswascollected:6,123usingescalatorsand4,896usinganadjacentstair.Analysisofthedatashowsthatthemajorityofpedestrians(67.0%)electedtousetheescalatorintheupdirection.However,inthedowndirectionthemajorityofpedestrians(65.3%)electedtousetheadjacentstair.Thissuggeststhatdirectionoftravelinfluencesescalator/stairselection.Anumberofevacuation/circulationmodelsthathaveanescalatorcapabilityutilise simplistic measures to determine if pedestrians elect to use an escalator or adjacent stair. In mostcases,pedestriansareeitherforcedtouseadevice,orselectadevicewhichminimisestraveltime.Basedonthedatacollected,clearlythisisnotalwaysthecase.Approximatelythreequartersofpedestrians(74.9%)rodetheescalatorswiththeremainingelectingtowalk.Ahigherproportionofwalkerswereobservedduringtherush-hourcomparedtothenon-rushhour.Thissuggeststhattrippurposeandsubsequentlevelsofmotivationinfluenceescalatorwalker/riderchoice.Theproportionofwalkers/riderswilldeterminetheflow-rateachieveduponanescalator.Aswithescalator/stairchoice,withoutthistypeofdata,manyevacuation/circulationmodelsignorethismicroscopicbehaviourandimposemacroscopicdatasuchasmaximalflow-ratesonescalatorperformance.Analysisofthesideusagedataidentifiedthattherewasacommonsidepreferenceforriderstotypicallyusetherightside(88.4%)andwalkerstousetheleftside(78.2%)ofeachescalator(p<0.05).Overallaveragewalkerspeedswerefasterinthedowndirection(0.82m/s)thanintheupdirection(0.70m/s)(p<0.05).

However,therewasnosignificantdifferencebetweenwalkerspeedsduringtherush-hourandnon-rushhourperiods.Unlikethewalker/riderchoice,thissuggeststhattrippurposeandsubsequentlevelsofmotivationareoflittleinfluenceuponescalatorwalkerspeeds.Themaximumescalatorflow-raterecordedwas75ped/minintheupdirectionduringthemorningrush-hourperiod.Thisisconsiderablylessthanmaximalflow-ratesreportedinpaststudies[2].

UsingthedatacollectedfromtheLondonunderground,aseriesof14undergroundstationevacuationscenarios were conducted exploring the influence of escalator strategies and associated human factors uponanevacuation.Theevacuationanalysisinvestigatedtheimpactofavarietyofescalatorhumanfactorsincludingdeviceselection,walker/riderselection,sideusage,andescalator/stairavailability.Thebuildingusedforthesimulationswasahypothetical2levelundergroundstation(platformandtickethalllevel)connectedviaaseriesofescalatorsandstairs.Intotal,3,856agentswererepresentedinthestation:2,892ontheplatformleveland964onthetickethalllevelabove.

Resultsfromtheanalysissuggestthateventheprovisionofstaticescalatorscanhaveaconsiderableinfluenceuponanevacuationcomparedtousingstairsalone.Itisnotedthatinsomecases,escalators–evenstaticescalators–arenotusedinevacuationsituations.Furthermore,theprovisionofamovingescalatorwasshowntodecreasetotalevacuationtimes(TET)byuptoapproximately25%comparedtousingstairsaloneandaround10%comparedtousingstaticescalators.ResultshaveshownthatlittledecreaseinTETwasobservedwhenallescalatoruserswalkedcomparedtoiftheyallrode.Assuch,urgingescalatoruserstorideonbothsidesofanescalator,maximisingtreadutilisation,maybeadvantageousduringanevacuationconsideringthereducedlikelihoodofescalatoruserstrippingcomparedtowalking.Duringscenarioswhereescalators/stairswererenderedunavailable,asexpected,hadaconsiderableimpactupontheevacuation.Inthosescenarios,increasesinTETofupto59.8%andplatformclearancetimesofupto71.6%comparedtostairsalonewererecorded.Suchfindingshighlighttheseveritycausedbytheunavailabilityofescalators/stairs(e.g.duetofire/smoke,codestipulations,etc),andtheneedtoconsideradditionalprovisionofverticalegresscapacity.

Thedataanalysisprovidesinsightintoescalatorhumanfactors.Itisausefulresourceforevacuation/circulationmodeldevelopers.Thepresentedevacuationanalysisdemonstratesthepotentialimpactofdifferentescalatorstrategiesandtowhatextenthumanfactorsinfluencethesestrategies.Thestudycouldpotentiallybeusedasabasisforengineeringanalysiswheretheperformanceofdifferentproceduralvariantswithregardstoescalatorsarerequiredtobeexplored.

REFERENCES

[1] Strakosch, G., Caporale, R.S, (2010), ‚The Vertical Transportation Handbook, The Third Edition‘, John Wiley and Sons Inc.

[2] Al-Sharif, L., (1996), ‚Escalator Handling Capacity: Standards Versus Practice‘, Elevator World.

[3] Cheung, C., Lam, W., (1998), ‚Pedestrian Route Choices Between Escalator and Stairway in MTR Stations‘, Journal of Transportation Engineering,. Vol. 124, No. 3, pp227-285.

[4] Davis, P., Dutta G., (2002), ‚Estimation of Capacity of Escalators in London Underground‘, No 2002-11-01, IIMA Working Papers from Indian Institute of Management Ahmedabad, Research and Publication Department.

[5] Kinsey, M.J, Galea, E.R, Lawrence, P.J, Blackshields, D., Hulse, L., Day, R. and Sharp, G., (2008), ‘Modelling Pedestrian Escalator Behaviour’, Pedestrian and Evacuation Dynamics (PED) Conference, pp689-695.

[6] Kinsey, M.J, Galea, E.R, Lawrence, P.J, (2009), ‘Extended Model of Pedestrian Escalator Behaviour Based on Data Collected within a Chinese Underground Station’, Proceedings of the Human Behaviour in Fire Conference, pp173-182.

[7] Kinsey, M.J, (2011), ‘Vertical Transport Evacuation Modelling’, PhD Thesis, University of Greenwich.

Experimental study of pedestrian dynamics

Cecile Appert-Rolland, University Paris-Sud, Paris FRANCEAsja Jelic, University Paris-Sud, Paris FRANCESamuel Lemercier, INRIA, Rennes FRANCEJulien Pettré, INRIA, Rennes FRANCE

IntheframeofaFrenchproject(PEDIGREE)involvingfourteams(IMT-Toulouse,INRIA-Rennes,CRCA-Toulouse,LPT-Orsay),wehaveperformedseveralexperimentstostudythedynamicsofpedestrians in various geometries.

Experimentswereperformedin-door,andeachpedestrianwasequippedwith4markerswhichweredetectedbyaVICONsystem(infra-redcameras).Asaresult,thefull3Dtrajectoryofeachpedestriancouldbereconstructed,withafrequencyof120frames/s.

Intheframeofthisproject,severalexperimentswereperformed,including:1-1d-motionalongacircle[1,2,3].2-mono-directionalandbi-directionalflowinaringcorridor[4].3-oscillationsatabottleneck.4-Interactionsbetween2or3individualtrajectories.5-Interactionsbetweenincominglinesofpedestrians.

Then,asanillustrationofthetypeofresultsthatcanbeobtainedfromtheseexperiments,weshallconcentrateontheanalysisofthe1dcircleexperiment.

Participantswereaskedtowalkinanaturalwayalongacircularpath,withoutpassingeachother.Twodifferentcircularpathswereused,withdifferentradii(respectively2.4and4.1meters).Thenumberofpedestrianswasvariedfrom8to28,resultinginaglobaldensityrangingfrom0.31to1.86ped/m.However,thelocaldensitiescouldbelowerorhigherduetospontaneousformationofspatialinhomogeneities.

Beforethestartoftheexperiment,pedestrianswereplacedeitheratequaldistancesaroundthecircle,orpackedalltogether.Eachexperimentlasted1mn(orslightlymoreforhigherdensities),andseveralreplicas(upto8)wererealizedforthesamesetofparameters,ifpossiblewithdifferentsetsofparticipants.

Firstsomecomparisonwithpreviousexperiments[5,6]willbepresented.Here,thefundamentaldiagramcanbedefinedinseveralways,dependingwhetherthedensityandthevelocityaretakenasglobal,locallyaveragedorinstantaneousquantities[1].Theinitialconditions(homogeneousorjammed)turnouttohaveashortlivedinfluence.Thecontributionstothefundamentaldiagramofstationary behavior or transient behavior due to rapid fluctuations will be discussed.

Inpreviousexperiments,thevelocity-spatialheadwayrelationwasfoundtobelinear.However,ourexperiment has allowed us to cover a much larger range of densities.Oneofourmainresultistoshowthatthereexisttwocleartransitionsinthebehaviorofpedestrians,separating3dynamicalphases[1]:-Freeflowregime:foraspatialheadwaylargerthan3meters,pedestrianswalkwiththeirpreferredvelocity;-Weaklyconstrainedregime:foraspatialheadwaybetween1.1and3meters,thevelocitydependsonlyweaklyonthespatialheadway.-Stronglyconstrainedregime:foraspatialheadwaylessthan1.1meter,thevelocitydependsstrongly on the spatial headway.Ineachphase,thevelocity-spatialheadwayrelationislinear,buttheslopeisdiffersfromonephasetotheother.Thisfindingcouldopennewperspectivesintermsofmodeling.

Thetransitionbetweentheweaklyandstronglyconstrainedregimeswasalreadypartiallyvisibleinpreviousexperimentsbutwasnotinterpretedassuch,aslowerdensitieswerenotexplored.Ithasconsequencesonthewaycomparisonsbetweendifferentexperimentsshouldbeperformed,inordertoextractsomeculturalinfluenceonthewalkingbehaviorasin[7].

Theoscillationsoftheradialcoordinateallowtodetectsteps.Weshallpresentsystematicallyhowthestepfrequencyoramplitudevarywiththeparametersoftheexperiment,andextractsomesimplelaws[2].Weshowalsothat,asexpected,synchronizationofthestepsmayoccurathighdensities,butthatthis is not a general feature.

Theknowledgeofthefullindividualtrajectoriesallowsalsotoextractthestopandgowavesthatareproducedathighdensities.Wecanextractnotonlythepropagationvelocityofthewaves,butalsothedampingofthesignalinsidethewave,thewidthofthewave,thecharacteristicsofthepedestriansinsidethewave,etc...Thepossibilitytolocalizepreciselythewaveswillenrichthediscussionaboutthefundamental diagram.

Theanalysisofthefollowingbehaviorobservedforpairsofsuccessivepedestrianshasledtotheproposalofamicroscopicmodel[3].Itispossibletocomparemodelsandexperimentseitheratthemicroscopiclevel,oratamacroscopiclevel,basedonthecharacteristicsofstopandgowaves.

Thedataobtainedfromthisexperimentwillbemadeavailableforthescientificcommunityonawebplatformin2012.

REFERENCES

[1] A. Jelic, C. Appert-Rolland, S. Lemercier and J. Pettre. Properties of pedestrians walking in line - the fundamental diagram, preprint, 2011.

[2] A. Jelic, S. Lemercier, C. Appert-Rolland, and J. Pettre. Properties of pedestrians walking in line - stepping behavior. In preparation, 2011.

[3] S. Lemercier, A. Jelic, J. Hua, J. Fehrenbach, P. Degond, C. Appert-Rolland, S. Donikian, and J. Pettre, A realistic model of following behavior for crowd simulation, submitted to Eurographics 2012.

[4] M. Moussaid, E. Guillot, M. Moreau, J. Fehrenbach, O. Chabiron, S. Lemercier, J. Pettre, C. Appert-Rolland, P. Degond, and G. Theraulaz. Outbreak and breakdowns of collective intelligence in human crowds. submitted to PLoS Computational Biology, 2011.

[5] A. Seyfried, B. Steffen, W. Klingsch, and M. Boltes. The fundamental diagram of pedestrian movement revisited. J. Stat. Mech., page P10002, 2005.

[6] A. Seyfried, B. Steffen, W. Klingsch, T. Lippert, and M. Boltes. Steps toward the fundamental diagram - empirical results and modelling. In N. Waldau, P. Gattermann, H. Knoflacher, and M. Schreckenberg (eds), Pedestrian and Evacuation Dynamics 2005. Berlin, Springer, 2007.

[7] U. Chattaraj, A. Seyfried, and P. Chakroborty. Comparison of pedestrian fundamental diagram across cultures. Advances in Complex Systems, 12:393-405, 2009.

Pedestrian agent based model suited to heterogeneous interactions overseen by perception

Laure Bourgois, IFSTTAR, Paris FRANCEJean-Michel Auberlet, IFSTTAR, Paris FRANCE

INTRODUCTIONHeterogeneouscrowdmodelsareneedinsomeapplicationsincludingeducation,entertainment,training(forthemilitaryandpolice)andhumanfactorsanalysisforbuildingevacuation,forsportevents[7]andformilitarysimulation(especiallyincaseofurbanwarfare[5]).

Mostoftheworkaboutpedestriansimulationsconcernsituationsofhomogeneousinteractionsasforplanningandfordesigningintransportationstudies.Thesestudiesassumethatpedestriansbroadlysharecommonbehaviourandgoal(reachingoutofabuilding,crossingastreet...).Exceptedinmilitarysimulationandgame,thereisonlyfewresearchwhereactorshaveantagonisticpurposes.Howevertheseresearchesdonotfocusspeciallyonelaboratingapedestrianmovingmodel,buttheyrathertreatofpatroloptimization(wheretosendefficientlythepolice…).

Inordertomodeldifferentinteractionsbetweenpedestriansincomplexanddynamiccontext/situation,weproposeanexploratoryworkaboutaflowofpedestriansinthepresenceofanimmediateandshiftingdanger(apacificdemonstrationwithamilitaryassault).WeproposehereanAgentbasedmodelwithahierarchicalarchitecturedrivenbyperception,articulatedwithacontrollersuitedforvarioussituations.Themodelisgenericsinceituseforpacificdemonstratorsandforcontrolforcemember.Weexperimentascenariowhereacontrolforcesquadtrytocatchsomeleadersinthecrowd.

STATEOFTHEARTMicroscopicmodelsrepresentpedestrianswiththemeanofindividualentities.Thesemodelsincludetwomajorapproaches:pedestriansubjectedtoforces[4]ortobehaviouralrules[8].Othermodelstryto encompass psychological features involved in high decision process of pedestrians.

WorksofMcKenzieaimtodevelopatoolwithasetofmodelsofmilitarysimulationwithacrediblecrowdpsychological[5].Crowd-MAGSproject[1]sharesthesameobjectives.WithaMultiAgentSystemapproach,eachpedestrianismodelledbyanagentendowedwithasetofcomplexbehaviors(set of hierarchical rules).

Recentworks[6]tackletheattractiveforcesbetweenpedestriansinorderreproducetheobservedrecurringpatternsingroups.Theseresearchesassumethatpedestriansareconsciousofthegrouprelationship.

Besides,precedentstudiesdonotcarrythemanypedestrianinteractionsinducedbyadynamicalenvironment.Indeed,onealterstheinitialinteractionbasedonhisperceptionoftheenvironment.

AHIERARCHIALARCHTIECTUREOurmodeltakesplacewithintheframeworkofperception-interpretation-selection-decision[10].Wenotethattheinterpretationandselectionprocesscanbedisconnectedandtheimportanceofdecisionfeedbackonthoseprocess.Wefollow[3]’sconceptthatdistinguish3decisionlevels:strategic,(planning…)tactical(sectionroutechoice…),operational.(instantaneousmotions…).Weproposeahybrid architecture driven by perception and we focus on the junction between tactical and operational level.Theoperationallevelmaybecontrolledindifferentwaysdependingonthenaturetasktoperform(crossingthestreet,escape...).

PEDESTRIANAGENTMODELBasically,weemploythewidelyusedSocialForceModelofHelbingmovingagents[4].Apedestrianagent acts as if it was subject to attractive forces (its destination ...) and repulsive forces (obstacles ...).Ittriestotrailstraightaheadinordertoreachitsdestinationfollowingashortestpathroute,andavoidingobstacles.Weenhancedthismodelwithsensitivitymeasuresandbyintroducingadditionalfeatures[2].

Weprovideherepedestrianagentswithdifferentfeaturesinordertoinfluencetheirperceivedinteractionandthewaytomanagethem.Inourexample,apacificdemonstratorhastoperceiveotherpacificpedestriansandobstaclesandtodifferentiatethemfromsomestationarydangerandfromsomechargingpolicemen;acontrolforcememberhastoperceivethepedestrian“target”andothernon-combatant pedestrians.

Formonitoringthemovingmodel,weproposeanautomatadrivenbyperception.Thistoolallowsforapedestriantoescapefromadanger(static,dynamicapredatorforinstance)andtogetcloserfromotherpacificdemonstrator.

DISCUSSIONANDPERSPECTIVESAftersomeexperimentations(forcalibratingsomeparametersofourmodel),weexhibittheemergenceofgroupphenomenon.Oursimulationstakeplaceinacorridorandinabottleneck.WeareabletocompareourresultswithworksliketheonesofVenel[9]andthoseofHoogendoorn[3].However,eveniftherearesomereferencesforcollectivebehaviour[11],wepointoutthelakeofdataaboutpedestrians.

More research is needed to better understand and to reproduce global behaviour for a group of pedestrians[6]onparticularinthecaseofcrowddispersioninanurbanwarfareorinanemergencyevacuation.Severalproblemsariselikegroupidentificationorgroupformation.

REFERENCES

[1] B. Moulin, B. Larochelle, Crowdmags, Multi-agent Geo-Simulation of the interactions of a crowd and control forces, Modelling, Simulation and Identification, 2010.

[2] L. Bourgois, A. Oulhaci, J.-M. Auberlet, Simulation de déplacement : Vers un modèle de perception et de prédiction du comportement d’autrui, JFSMA, 2010.

[3] S. P. Hoogendoorn, W. Daamen, Pedestrian Behavior at Bottlenecks, Journal Transportation Science, 2005.

[4] A. Johansson, D. Helbing, P.K. Shukla, Specification of the Social Force Pedestrian Model by Evolutionary Adjustment to Video Tracking Data, Advances in Complex Systems, 2007.

[5] F. McKenzie, M. Petty, P. Kruszewski, R. Gaskins Q.-A. Nguyen, J. Seevinck, E. Weisel Integrating crowd-behavior modeling into military simulation using game technology, Journal Simulation and Gaming, March 2008

[6] M. Moussaïd, N. Perozo, S. Garnier, D. Helbing, G. Theraulaz, The walking behaviour of pedestrian social groups and its impact on crowd dynamics, PLoS ONE, 2010.

[7] N. Pelechano, K. O’Brien, B. Silverman, N. Badler, Crowd Simulation Incorporating Agent Psychological Models, Roles and Communication, 1st Int‘l Workshop on Crowd Simulation, 2005

[8] C.W. Reynolds, Flocks, Herds, and Schools: A Distributed Behavioral Model, Computer Graphics, SIGRAPH, 1987.

[9] J. Venel, Integrating strategies in numerical modeling of crowd motion, Pedestrian and EvacuationDy namics, 2008.

[10] D. Weyns, E. Steegmans, and T. Holvoet. Towards active perception in situated multi-agent systems, Applied Artificial Intelligence, 2004.

[11] U. Weidmann, Transporttechnik der Fussgänger, Transporttechnische Eigenschaften des Fussgängerverkehrs (Literturauswertung) Schriftenreihe des IVT Nr. 90, Zweite, ergänzte Auflage, Zürich, März, 1993.

Measuring individual’s egress preference in wayfinding through virtual navigation experiments

Jan Dijkstra, Eindhoven University of Technology, Eindhoven NETHERLANDSQunli Chen, Eindhoven University of Technology, Eindhoven NETHERLANDSBauke de Vries, Eindhoven University of Technology, Eindhoven NETHERLANDSJoran Jessurun, Eindhoven University of Technology, Eindhoven NETHERLANDS

Finding one’s way to a certain destination is one of the most compulsive behaviours in our daily life. Thishasbeenextensivelystudiedinthelast50years.OriginallyraisedbyLynch(1960),theterm“wayfinding”isdefinedbyGolledge(1999)as“theprocessofdeterminingandfollowingapathorroutebetweenanoriginandadestination.”ArthurandPassini(1990)suggestthatdifferenttypesofinformationcanbeusedinwayfindingtasks,namelytheverbal(informationobtainedfromthereception,staffmembers,etc.),thegraphic(mapoftheenvironment,signageshowingthelocationorpointingtoacertainlocation,etc.),thearchitectural(e.g.doors,corridors,texture,elevators)(DoguandErkip2000,Passini1996),andthespatial(spatialrelationshipofobjectsintheenvironment)(ArthurandPassini1992,Passini1984).Thesignificancesofthesetypesofinformationonwayfindingprocesseshavebeeninvestigated.Inmoststudies,however,thefunctionofthearchitecturalinformationwasunderestimatedcomparedtoothertypesofinformation--itwasoftentreatedaslimiting the architectural space.

Inreality,humanfeelingsaboutthecolourofanobjectareinfluencedbydifferentfactors,suchasthephysicalconditionoftheobject,thecolourcontrastbetweentheobjectanditsbackground,ourmentalstateatthatmoment,etc.Moreover,differentcolourshaveadifferentpsychologicaleffectfordifferentpeople.Inthisstudy,agreyscalemodelisappliedfordifferentiatingdifferentcoloursandtodescribevariouspsychologicaleffectsofcoloursbythisgreyscalemodel.Therefore,egresschoicesareonlyinfluencedbygreycontrastbetweenegresses,andbetweentheegressanditsbackground.Thisgreycontrast represents the colour factor.

Inthispaper,astudywillbepresentedwheresubjects’egresschoicesarecollectedgivendifferentsettings of selected egresses (door or corridor) in a list of architectural spaces. In this context architecturalspacesareindependentspaceswithonlyvisualarchitecturalinformationprovided.Themainobjectiveofthisstudyistoresearchtheutilityeffectofvariablesthatwouldinfluenceindividual’segresschoicesinarchitecturalspaces.Moreover,itisassumedthatapreferencefunctioncanbededucedtopredicttheprobabilityofanindividualtochoosetheleftegressoftheegressespair(left,right)inarchitecturalspaces.Asuggestionforthisdeductionwillbedone.

Thispaperfocusesontheindividual’segresschoosingbehaviourinthebuiltofficeenvironment.Notethat“egress”mentionedinthispaperisspecifiedasalinkagebetweenrooms,oralinkagebetweenaroomandaspace,ratherthanthelinkagebetweenabuiltenvironmentandtheworldoutofthisbuiltenvironment.Afterastudyontheegresstypesintheofficeenvironment,twocommonusedegressesareselectedfortheexperimentaldesign:doorandcorridor.Inaddition,threedifferentsizesaredesignatedforeachassignedegressrefereedtoasnarrow(80cmby200cm),normal(150cmby200cm),

andlarge(194cmby246cm).

Fourcharacteristicsareidentifiedforthesetwoegresstypes,whichcouldinfluencevisitor’schoiceontheegresses,resultinginthedeductionoffourvariablesfordesigningtheroomsintheexperiment:thedistancefromtheegresstotheentrancepoint(D),thewidthoftheegress(W),thegreyleveloftheegress(G),andtheanglebetweentheviewdirectionandtheegressintheview(A).Thegreylevelofanegressiscomputationallydefined,rangingfrom0to100,indicatingthebrightnessofthegreyoftheegress.Thechangeofroomilluminationinthevirtualbuiltenvironmentwouldaffectthegreyleveloftheegresses.Thehigherthegreylevelofanegress,thebrighteroftheegress’sgreywillbe.

Thetypeofroomisanotherrelatedvariable,whichisalsotakenintoaccountintheexperimentaldesignphase.Intheexperiment,eachdesignedroomconsistsoftwoegresses.Asaresult,therearethreetypesofroomlayouts:aroomwithtwocorridors,aroomwithtwodoors,andaroomwithadoor and a corridor.

Thedesignedroomsareofthreedifferentsizes,namelysmall(4mby6m),medium(9mby12m),andlarge(15mby20m).Eachdesignedroomcontainstwoegresses:oneontheleft,andoneontheright.Theorientationsoftheegressesintheseroomsvarysystematically.TheexperimentisVRapplicationbased.Thereforetheaspectratioofthescreenispresetsuchthattheapplicationcouldberunonanycomputerwiththesameaspectratio.Basedonthevariablelevelsoftheegressesandtypesoftherooms,fractionalfactorialdesignisemployedtodeterminethenumberofconvexroomsthatisrequiredfortheexperiment.Thesequentialorderoftheroomspresentedintheexperimentisrandomlygeneratedforeachsubjectoncetheystarttheexperiment.Inthisexperiment,aslongasonesubjectchoosesanegressinthecurrentroomandwithinacertaindistancetothechosenegress,thenextroomwouldbegeneratedandlinkedtothechosenegressautomaticallyaccordingtothegeneratedsequentialorderoftheroomsforthissubject.

Fromtheanalysisonecanconcludethatvisualarchitecturalinformationhasasignificantutilityeffectonindividuals’behaviours.Thispaperwillpresenttheexperimentaldesign,thedatacollection,andtheoutcomesoftheutilityeffectsaswellasthesuggestionfortheutilityfunction.

Oral presentations


A.4 Edwin Galea, University of Greenwich, London UNITED KINGDOM An evacuation validation data set for large passenger ships

A.5 Hermann Mayer, Siemens AG, München GERMANY Influence of emissions on pedestrian evacuation

A.6 Christian Rogsch, RiMEA e.V., Neustadt GERMANY RiMEA - A way to define a standard for evacuation calculations

B.4 Mojdeh Nasir, Deakin University, Melbourne AUSTRALIA Fuzzy prediction of pedestrian steering behaviour with local environmental effects

B.5 Ekaterina Kirik, Institute of Computational Modelling RAS, Novosibirsk RUSSIA Fundamental diagram as a model input – direct movement equation of pedestrian dynamics

B.6 Francesco Zanlungo, Advanced Telecommunication Research Institute, Kyoto JAPAN Experimental study and modelisation of pedestrian space occupation and motion pattern in a real world environment

C.4 Laura Künzer, Friedrich-Schiller-Universität Jena, Jena GERMANY Psychological aspects of german signal words in evacuation warnings

C.5 Tomoichi Takahashi, Meiji University, Tokyo JAPAN Effect of guidance information and human relations among agents on crowd evacuation behavior

C.6 Felix Cabrera, Pontificia Universidad Catolica del Peru, Lima PERU Using the social force model to represent the behavior of pedestrians at chaotic intersections of developing countries: the case of Peru

An evacuation validation data set for large passenger ships

Edwin Galea, University of Greenwich, London UNITED KINGDOMSteven Deere, University of Greenwich, London UNITED KINGDOMRobert Brown, Memorial University of Newfoundland, St. John‘s NL CANADALazaros Filippidis, Univesity of Greenwich, London UNITED KINGDOM

In2002theInternationalMaritimeOrganisation(IMO)introducedguidelinesforundertakingfull-scaleevacuationanalysisoflargepassengershipsusingshipevacuationmodels[1].Theseguidelines,knownasIMOMSCCirc.1033,weretobeusedtocertifythatthepassengershipdesignwasappropriateforfull-scaleevacuation.Aspartoftheseguidelinesitwasidentifiedthatappropriatefull-scaleshipbasedevacuationvalidationdatawasnotavailabletoassessthesuitabilityofshipevacuationmodels.Assuitablevalidationdatawasnotavailable,aseriesoftestcasesweredevelopedwhichverifiedthecapabilityofproposedshipevacuationmodelsinundertakingsimplesimulations.However,theseverificationcaseswerenotbasedonexperimentaldata.Furthermore,successfullyundertakingtheseverificationcasesdoesnotimplythattheevacuationmodelisvalidatedorcapableofpredictingrealevacuationperformance.In2007IMOMSCCirc.1238[2],amodifiedsetofprotocolsforpassengershipevacuationanalysisandcertificationwerereleasedhowever,theissueofvalidationofpassengershipevacuationmodelswasnotaddressed.TheIMOFireProtection(FP)Sub-CommitteeintheirmodificationofMSCCirc.1033attheFP51meetinginFebruary2007[3]invitedmembergovernmentstoprovide,“…furtherinformationonadditionalscenariosforevacuationanalysisandfull scale data to be used for validation and calibration purposes of the draft revised interim guideline.” TheEUframework7projectSAFEGUARDaimstoaddressthisrequirementbyprovidingfull-scaledata for calibration and validation of ship based evacuation models.

AspartofprojectSAFEGUARD,fivepassengerresponsetimedatasetsandfivefull-scalevalidationdatasetsfromthreedifferenttypesofpassengervesselshavebeencollected.ThispaperwillconcentrateonthefirstdatasetcollectedfromthelargeRO-PAXferryoperatedbyColorLineAScalledSuperSpeed1[4].Thevesselcancarryapproximately2000passengersandcrewandover700vehicles.TheroutetakenbythevesselisfromLarvikinNorwaytoHirtshalsinDenmark,atripof3hoursand45minutes.Theshipcontainsamixtureofspacesspreadoverthreedecksincluding;businessandtravellerclassseatingareas(airlinestyleseating),largeretailandrestaurant/cateringareas,barareas,in-doorandout-doorgeneralseatingareasandgeneralcirculationspaces.DatafromasailingfromLarviktoNorwayinearlySeptember2009wascollectedwith1349passengersonboard.Thetrialconsistedoftheship‘sCaptainsoundingthealarmandcrewmovingthepassengersintothedesignatedassemblyareas.Thetrialtookplaceatanunspecifiedtimeonthecrossinghowever,passengerswereawarethatontheircrossinganassemblyexercisewouldtakeplace.Thedatacollectedduringtheassembly trial consisted of passenger response time data and a comprehensive validation data set forshipbasedevacuationmodels.Some30digitalvideocameraswereusedtocollecttheresponsetimedata.Thevalidationdatawascollectedusinganoveldataacquisitionsystemconsistingofship-mountedbeacons,eachemittinguniqueInfra-Red(IR)signalsanddataloggingtagswornbyeach

passenger[4].Thevalidationdatasetconsistsofstartinglocationsofpassengers,arrivaltimeatthedesignatedassemblylocationsandthepathstakenbythepassengers.

ThispaperwillpresenttheresultsgeneratedbytheIRloggingsystemandsomevalidationanalysisundertakentodemonstratethattheIRsystemprovidesanaccuraterepresentationofthearrivaltimesattheassemblystations.ThisisbasedoncomparisonsofmeasuredIRarrivaltimeswiththatofarrivaltimesmeasuredusingvideodata.TheresultsofthisanalysisconfirmthattheIRloggingsystemcanprovide a very accurate measure of arrival times for large crowds of people.

Theresultoftheassemblyexerciseisanassemblycurvewhichplotsthearrivaltimeofeachpassengerintotheassemblystations.Assuchtheevacuationmodelvalidationexercisewillmeasuretheabilityoftheevacuationmodeltopredicttheoverallassemblycurve–notsimplytheoverallassemblytime.Itisnotedthatforthisexercise,thetotalassemblytimewasapproximately10minutes.Inprinciplethisdatasetisidealforvalidationpurposes,asthestartinglocationsandresponsetimesofeachparticipantisknown.Thismeansthatitshouldbepossibletoremovemostoftheuncertaintyassociatedwithinput parameters associated with response time and starting location.

Unfortunately,thesituationwasnotasidealasenvisaged.First,ofthe1349passengersonboard,780woretheIRtagsandparticipatedintheassemblytrial.Themajorityofthe569passengerswhodidnottakethetagsindicatedthattheydidnotwanttoparticipateintheassemblyexercise–whichwasnotcompulsoryforethicalandlegalreasons.However,ofthe569passengerswhodidnottaketags,asignificantnumber(asdeterminedfromtheanalysisofvideofootage)dideventuallydecidetoparticipateintheassemblyexercise.Byparticipatinginthetrial,thepresenceoftheuntaggedindividualsintheevacuationrouteswillhavehadanimpactontheoverallevacuation,especiallyinthehighlycongestedareas.However,theirassemblytimeswillnothavebeenrecordedintheoverallassemblycurve.Secondly,theexactlocationofthetaggedpeoplewasnotknown,buttheregionthattheywerelocatedinwasknown.Spatialregionswherebetween24mand48mlong;thusnotknowingtheprecisestartinglocationofanindividualmayhaveincreased/decreasedtheirarrivaltimeby25-50seconds.Furthermore,theresponsetimeisnotassociatedwithauniqueindividualbuttoaregion.Thusthepreciseresponsetimeofeachuniqueindividualisnotknown,buttheresponsetimedistributionassociatedwithastartingregionisknown.Allofthesefactorsmustbetakenintoconsiderationwhendetermininghowwellanevacuationmodelpredictstheassemblyexercise.Thesefactors and how they are addressed are described and discussed in the paper.

Finally,theresultsfromblindsimulationsusingmaritimeEXODUSforthisassemblyexercisearepresentedandcomparedwiththemeasureddata.Anumberofmeasuresareproposedtoassessthegoodnessoffitbetweenthepredictedmodeldataandthemeasureddata,includingthosepresentedin[5].Inaddition,amethodologyisproposedforrunningvalidationsimulationsusingthisdatasetwhich includes a set of appropriate acceptance criteria. It is proposed that these protocols could be usedbyIMOaspartofavalidationsuitetodetermineacceptabilityofmaritimeevacuationmodels.

ACKNOWLEDGEMENTProjectSAFEGUARD(contract218493)isfundedundertheEuropeanUnionFramework7Transportinitiative.Theauthorsacknowledgetheco-operationoftheirprojectpartners.

REFERENCES

[1] IMO, “Interim Guidelines for Evacuation Analyses for New and Existing Passenger Ships”, IMO MSC/Circ 1033, 6 June 2002

[2] IMO MSC Circ 1238, 2007

[3] FP 51/WP.3, 8 Feb 2007

[4] Galea, E.R., Brown, R.C., Filippidis. L., and Deere. S., “Collection of Evacuation Data for Large Passenger Vessels at Sea”, Pedestrian and Evacuation Dynamics 2010. 5th International Conference. Proceedings. March 8-10, 2010, Springer, New York, NY, Peacock, R.D., Kuligowski, E.D., and Averill, J.D., Editor(s), pp 163-172, 2011

[5] Peacock, R.D., Reneke, P.A., Davis, W.D., Jones, W.W., “Quantifying Fire Model Evaluation Using Functional Analysis, Fire Safety Journal, 22 (1999), pp167-184

Influence of emissions on pedestrian evacuation

Hermann Mayer, Siemens AG, München GERMANYDirk Hartmann, Siemens AG, München GERMANYOliver Zechlin, Siemens AG, München GERMANYWolfram Klein, Siemens AG, München GERMANY

INTRODUCTIONWehavedevelopedapedestriansimulation,whichiscapableofintegratinganemissionmodelthatinfluencesthemovementofoccupants.Twoalternatives,onlineandofflinecouplinghavebeenimplemented.Offlinecouplingcanbeemployedtointegratedatageneratedbyaccurate,buttime-consumingmethods,likeNISTFDS(FireDynamicsSimulatordevelopedbytheNationalInstituteofStandardsandTechnology).ThedistributionofsubstancesiscalculatedbyaNISTFDSandafterwardstheresultsaresynchronizedwiththedownstreampedestriansimulation(incontrasttodirectcoupling[1]).Foronlinecouplingtheemissionisnotpre-calculatedbyastandalonemodel,butdirectlyintegratedintothemainloopofthepedestriansimulation.Therefore,animmediateinteractionbetweenoccupantsandtheeffectsofacertainemissionisfeasible.Inaddition,theappliedalgorithmsareoptimizedforperformance,whilestillprovidingacceptableaccuracy.Thisfeatureisessentialforonlineegressplanningasdemandedbyfirealarmsystems:bothfirespreadandpedestrianmotioncanbesimulatedfasterthanreal-time.Incaseofafire,thesimulationcalculatessituational,individualegressroutesforeachoccupant(oreachgroupofoccupants).Usingsuper-real-timecalculation,theupdated egress route can be communicated to the occupants immediately.

MATERIALSANDMETHODSForaccurateofflinesimulation,fireandegresscomputationsareperformedseparatelyandsequentially.First,thepropagationofafireissimulatedbyadedicatedfiremodel(wehaveintegratedNISTFDS,whichisbasedoncomputationalfluid/gasdynamics:CFD).Afterwardstheresultofthefiresimulationispipedintoamodelforpedestriansimulation.Thetemporaldistributionofthefireismodeledbypotentialfields,repellingpedestrians.However,sincethedistributionofthefireiscomputedseparately,aninteractionbetweenthepedestriansandthefirecanhardlybemodelled.Therefore,weadditionallyprovideonlinecouplingoffireandegresssimulation,i.e.aftereachtimestepofthefiresimulationthepedestriansimulationisupdatedandviceversa.DuetothecomputationalcomplexityofmodelsbasedonCDF,thoseapproachescannotbeperformedinreal-time,whichhoweverisastrictrequirementforincorporatingthesimulationintoafirealarmsystem.Real-timesimulationisamustfortimecriticalapplicationslikevoiceevacuationanddynamicegressroutes:an optimal egress plan is calculated by a simulation and communicated to the occupants online. In addition,thehighaccuracyofafiremodelbasedoncomputationalfluiddynamicsisnotnecessarilyrequiredforthecontrollerofsuchafirealarmsystem.Thereforewehavedevelopedaseparatemoduleforonlinefiresimulation(seebelow).

Sincewehavechosenacellularautomatonforthemovementsofpedestrians,ourapproachofintegratingafiresimulation(offlineandonline)isalsobasedonthismodel.Theexaminedareais

discretizedintohexagonalcells,whereeachcellcanbeoccupiedbyoneperson.Ifcellsareafflictedbythedistributionofsmokeorfire,thosecellswillbemarkedasblockersforthenextstepinpedestriansimulation.Thismightcontingentlyresultinarecalculationofthebestegresspathforsomepedestrians.

WhilethefiredistributionoftheofflinecouplingisderivedfromNISTFDS,wehavedevelopedadedicatedmodelforonlinecouplingofemissions(likefireandsmoke).Anefficientmethodtocoupleacontinuousmodelwithadiscretizedsimulationenvironmentisconstitutedbytheso-calledEikonalequation.TheEikonalequationisusedtocalculatethearrivaltimeofwaves,whicharepropagatedbyacertainspeed.Thespeedofwavepropagationisnotlimitedtoaconstantvalue,butcantakeanyfunctiondependingonthepositioninthesimulationarea.Possibleimplementationscompriseadependencyofthewavepropagationonmaterialparameters,oronexternalparameterslikethewindspeed,oronelevationprofilesorwhatevermightinfluencethedistributionofafire.

Apartfromtheadvantagesmentionedabove,itisalsopossibletocoupleasimulationwithmorethanoneEikonalequation,forexampleifdifferentmodelsforthepropagationoffireandsmokearerequired.Furthermore,obstaclescanbeincludedintothesimulationbysimplysettingthespeedofwavepropagationtozeroinsidetheseareas.AlthoughthesimulationbasedonanEikonalequationisacontinuousmethod,thenumericalsolutionoftheequationissubjecttoacertaindiscretizationonacomputer.However,afinediscretizationcanbechosen,sinceefficientmethodsareprovidedtosolveEikonalequations(e.g.theso-calledFastMarchingMethod[2],analgorithmwhichsolvestheEikonalequationinO(nlog(n))steps,wherenisthenumberofpointsinthediscretization).Giventheseprerequisites,thedistributionoffireandsmoke(andotherharmfulsubstances)canbecoupledtoareal-timepedestriansimulation.Inaddition,thosesimulationmodelsprovideanaccuracysufficientforonlineevacuationapplications.However,thehighaccuracyoffine-griddedCFD-basedmodelslikeNISTFDScannotbereached,i.e.theofflinemodeisstill the favorable option for detailed planning purposes.

EXAMPLESInoneexamplewehavesimulatedthespreadoftoxicsubstances(likesmoke)intheentranceareaofasubwaystation.Therearetwoalternativeegresspaths(passagewaysleadingfromthesubwayexittothemainescalator/exitstair).Oncethetoxicsubstancespreads,theshortestegresspathiseventuallyblockedafteragiventime.Therefore,thepedestriansareforcedtotakeanalternativeegresspath,whichislongerbutnotaffectedbytoxicsubstances.Thiskindofsimulationcanbeemployedtotestthelayoutofegresswaysunderextraordinaryevents(likefireandsmokeemission).

Asecondexampleassessestheevacuationofoccupantsfromamulti-storeybuilding.Thegoalofthesimulationisfindinganoptimumegressrouteforeachoftheoccupants(i.e.thesimulationcalculatestheoptimumegressrouteforeachindividualperson).Theoptimumegressroutemightdifferfrompersontoperson,sincetheoccupantsarelocatedindifferentpartsofthebuildingwithdifferentaccessibilityoftheexits.Thiskindofsimulationcouldbedirectlylinkedtoanintelligentfirealarmingsystem,whichadviceseachindividualoccupanthowtoexitthebuildingincaseofanemergency.Possiblechannelsofnotificationare,forexample,sendingashortmessageonthemobile,displayingtheexitrouteontheindividualdesktop,orusingvoiceevacuationetc.Inordertocalculatesuchdetailedplans,thesimulationneedstogatherasmuchinformationaspossibleabouttheenvironment.Therefore,regardingthisexample,thefirealarmingsystemhastobeconnectedtoavarietyofsensors,likeoccupancydetection(e.g.cameras,sensitivefloorplates)orfire/smokedetection.

REFERENCES

[1] Korhonen, T. and Hostikka, S. (2009). Fire Dynamics Simulator with Evacuation: FDS+Evac. Technical Reference and User’s Guide. VTT Technical Research Centre of Finland.

[2] Sethian J. (1999). Level Set Methods and Fast Marching Methods. Cambridge University Press.

RiMEA - A way to define a standard for evacuation calculations

Christian Rogsch, RiMEA e.V., Neustadt GERMANYHubert Klüpfel, RiMEA e.V., Neustadt GERMANYRainer Könnecke, RiMEA e.V., Neustadt GERMANYAndreas Winkens, RiMEA e.V., Neustadt GERMANY

SincemanyyearstheRiMEA-Richtlinie(RichtliniefürMikroskopischeEvakuierungsAnalysen-GuidelineforMicroscopicEvacuationAnalyses;RiMEA-Guideline)isaguidelineforGerman-speakingauthoritiestocheckevacuationanalysesofcomplexbuildings.BasedontheRiMEA-Guidelineexpertreportsarewrittentoensurethatthefundamentalquestionsofanevacuationanalysisare answered.

SinceNovember22nd,2011,aGermanDINSpecification(likeaPre-Standard)forevacuationcalculations(DINSpec91284)isavailable.ThisDINSpecificationisbasedontheRiMEA-GuidelineandwasdevelopedtogetherwiththeRiMEAsociety(RiMEAe.V.).ThefirstpartofthispresentationwillshowhowthisDINSpecification(DINSpec91284)wasdevelopedandwhatkindofrequirementshavetobefulfilledto“transform”theRiMEA-GuidelinetothisGermanDINSpecification.AdditionallywewillshowwhatarethedifferencesbetweenthisDINSpecificationandtheRiMEA-Guideline.

InanAnnextheRiMEA-Guidelinepresentsasetoftestcasesformicroscopicsoftwaretoolsforevacuationcalculation.Thiswillbethesecondpartofthepresentation.Thesetestcasesshouldbeusedbythesoftwaredevelopertoshowresultsofthesoftwarebythemselves,andtopresentresultsinawrittenreporttoe.g.localauthoritiesforvalidationofthemodelusedinthesoftwareandverficationof its correct implementation.

Thesetestcasesareverybasic,toensurethatanysuitiblesoftwareisabletorunthetestscenarios.Thegoalofthesetestcasesistoshowusersofsoftwaretoolsthedifferencesorequityofdifferentsoftwaretools,furthermorethesetestcasesshouldshowlocalauthoritiesthatthesoftwaretoolswhichhavebeenusedforanevacuationreportfulfillatleastthisminimumstandardforsoftwaretools.Atthemoment(November2011)theRiMEA-Guidelineconsistsof14testcasesforsoftwaretools.Thesetestcasesarecombinedtodifferentgroupsofverificationandvalidationtestcases,thuswehave-Testcasesfordifferentsoftwarecomponents(testcase1–7)-Functionalverification/validation(testcase8)-Qualitativeverification/validation(testcase9-14)

In the following we present a short overview about these test cases:-Testcase1:Keepvelocityinafloor-Testcase2:Keepvelocityupstairs-Testcase3:Keepvelocitydownstairs-Testcase4:Specificflowthroughanopening

-Testcase5:Responsetime-Testcase6:Movingaroundacorner-Testcase7:Assignmentofdemographicparameter-Testcase8:Parameterstudy-Testcase9:Evacuationofalargeroom-Testcase10:Assignmentofevacuationroutes-Testcase11:Choosinganevacuationroute-Testcase12:Effectsbasedonbottlenecks-Testcase13:Congestioninfrontofstairs-Testcase14:Routechoice

Inthispartofthepresentationwewillshowresultsofdifferentsoftwaretoolsbasedonthetestcases,whicharepartoftheRiMEA-Guideline.Thegoalofthispartofthepresentationisalsotoassuredevelopersofinternationalnon-Germansoftwaretoolstousethiskindofverification/validationsuitetoshowthecapabilitiesoftheirsoftwaretools.FurthermorewewouldliketoinviteeverybodytosubmitnewtestcasesfortheRiMEA-Guideline,thusthenumberoftestcaseswillincrease,thusattheendawideanwellaccepted„benchmarksuite“canbepresentedtousersofsoftwaretoolsforevacuationcalculationandtolocalauthorities,whichhavetoacceptresultswhicharecalculatedbythese software tools.

InthelastpartofthepresentationwewillshowhowtheRiMEAguidelinewasdevelopedoverseveralyears.ThereforewewillgiveanhistoricaloverviewaboutthedevelopmentoftheRiMEAguideline,whichwasfirstpresentedatPED2005inViennatoaninternationalaudience.BasedonthefirstversionoftheRiMEA-Guideline(January2004)andtheactualedition,whichwillbeadoptedattheRiMEAmembermeetinginMarch2012wewillshowhowtheguidelinegrows(18pages2004,actually32pages2011)andwhatchangeshavebeendone.WefurthermorewanttoshowhowanewversionoftheRiMEA-GuidelineisdevelopedandwhatarethefurtherplansofRiMEAe.V.,thesocietywhichisresponsibleforthisGuideline.

Fuzzy prediction of pedestrian steering behaviour with local environmental effects

Mojdeh Nasir, Deakin University, Melbourne AUSTRALIANahavandi Saeid, Deakin University, Melbourne AUSTRALIADouglas Creighton, Deakin University, Melbourne AUSTRALIA

ABSTRACTThisresearchfocusesonpredictionofpedestrianwalkingpathsinindoorpublicenvironmentsduringnormalandnon-panicsituations.Theaimistoincorporateimpreciseandsubjectiveaspectsofpedestrianinteractionwiththeenvironmenttoenhancesteeringbehaviourmodelling.Theproposedmodelintroducesafuzzylogicframeworktopredicttheimpactofenvironmentalstimuli,withinapedestrian’sfieldofview,onmovementdirection.Attractiveandrepulsiveeffectsofthesurroundingenvironmentarequantifiedbysocialforcemethod.Ahighflowcorridorinanofficeareaisconsideredasacasestudy.Stochasticsimulationisusedtogeneratewalkingtrajectories,calculateadynamiccontourmapofenvironmentalstimuliineachstepandrecognisethehighflowwalkableareasinthecorridor. FUZZYLOGICSTEERINGPREDICTIONMODELWayfindingbehaviouralstudieshaverevealedthatthespatialabilityofpedestriansallowsthemtofindapathfromthecurrentlocationtoadestination[1].Duringwayfindingactivitiespedestriansareconfrontwithenvironmentalstimulationsthatchangedynamicallyaftereachstep.Environmentalstimulationshaveimportantinfluenceonvisuallydirectedwalkingtasks.However,variablefactorssuchastripintentionandpedestrian’sattributesarecontributingelementsthatmakethepredictionofpedestrian–environmentinteractionsanimpreciseandfuzzyproblem.

Dynamicalchangesofenvironmentalstimulationsconstantlyupdatethepedestrians’worldviewoftheirsurroundingsandaffecttheirperception.Understandingapedestrian’sperceptionofenvironmentalstimuliisnecessarytoaccuratelyestimatethepedestrianmovement[2].Itisbelievedthat information exchange within a dynamic environment contributes to the control of human locomotiontasks[3].Golledgeetal.[4]elevatedthequestionofinformationintegrationoftherouteattributes.Theyhighlightedthatorientationandmovementdirectionaremorefuzzyandrelatedtoawiderangeofelements.Recently,fuzzyrule-basedsystemshavealsobeensuccessfullyappliedinthefieldofrobotnavigationandpathplanning[5].Fuzzylogichasthecapabilitytomodeltheimpreciseand diverse nature of pedestrian perception and reaction towards the environmental impacts.

Thisstudyhighlightstwochallengingissues:firstly,itaddresseshowtoquantifytheenvironmentaleffects,andsecondly,howtheenvironmentalstimuliinfluencethespatialbehaviour.Inthiscontext,wehaveproposedafuzzylogicapproachtomodelthelocalsteeringbehaviour.Thefuzzysystemcomprisesofthreeinputs,oneoutputand216rulesasillustratedinFigure1.Inputsaretheagent’sperceptionfromthreepossiblefuturetravelpoints,whicharedescribedas{FrontPosition,RightPosition,LeftPosition}.Thesummationofattractiveandrepulsivestimulationisassessedforeach

futureposition.Thisperceivedinformationisthenrecognisedas{Highattractive,Mediumattractive,Lowattractive,Lowrepulsive,Mediumrepulsive,Highrepulsive}toexpresslevelofperception.Therefore,eachinputconsistsofsixmembershipfunctionsandchangeofmovementdirectioncanbeinferredbytheoutputoffuzzysystem.

Tosuccinctlyexpresspedestrianbehaviour,wehavemadethefollowingfiveassumptions:(i)alltheobjectsinduceattractiveorrepulsiveeffectstothesurroundingenvironment,andthecumulativesumofattractiveandrepulsivestimulationsiscomputedforeachpointintheterrainsurface;(ii)thelevelsofstimulationinthefront,rightandlefthandsideoftheagentaretheinputsofthefuzzysystem;(iii)ineachstepalongthemovementpath,threealternativesexist(moveforward,changedirectiontotherightortotheleft);(iv)theagentisabletochangethedirectionbetweenacontinuousrangefrom,-12to+12degrees,insubsequentsteps;and(v)theangularchangeofdirectionisthefinalcommandformovementdrivenfromfuzzyrules.Thethreefundamentalelementsofthemodelinvestigatedarepsycho-sociologicalmotivations,dynamicenvironmentalinformationthatisprovidedbyobjectslocatedinthefieldofviewtoreflectlocalawareness,andthemovementdirectionthatistheoutputoffuzzylogicsystem.

SimulationoffuzzysteeringmodelwithlocalenvironmentaleffectsContradictorypsycho-sociologicalforcesmotivatetheagenttomovetowardsadesireddestination.TheHelbingsocialforcemodelisoneofthemorepracticalandreliablemethodsdescribingthebehaviourofpedestrians,whichconsiderstheeffectsofattractiveandrepulsiveforces[6].Wehaveadoptedthisapproachtoquantifytheenvironmentalinfluences.Figure2showsthetraceofwalkingpathsinthesimulatedcorridorandthepotentialattractive,repulsiveandtotalstimulationinthatareaduetotheprinterandtheexitdoor.Itisassumedthattheprinterhasbothattractiveandrepulsiveeffects,whilethe exit door provides an attractive influence. Further simulations have been completed to gain a better understanding of algorithm performance and the impact of dynamic changes of environmental stimuli.

CONCLUSIONThesimulationresultsindicatethattheproposedfuzzy-basedapproachisapromisingmethodtomodelthepedestrianwalkingpathundernormalconditions.Inthismethodology,theindividual-basedrepresentationofterrainemploystheconceptoffieldofviewtocapturevisualstimulithatimpactwalkingdirectionandacquiredynamicenvironmentalinformation,whichisanecessarycharacteristic for local awareness.Toverifytheconcept,atwodimensionalspacewithwalls,printer,entranceandexitwasstudied.Aspedestrianinteractionwiththeenvironmentisanimportantfeatureofsteeringbehaviour,weassessedthelevelofinducedeffectsexertedbytheenvironmentonthepedestrianemployingsocialforcemodeltoobtaintheturningangleofdirectionforthenextstepusingfuzzylogicframework.Inthisregard,fuzzylogicoffersaframeworktomodeltheproblemconsideringbothimprecisenatureofenvironmentaleffectsanddiversityinpedestrianperceptionanddecisionmaking.

REFERENCES

[1] D. Gibson, The Wayfinding Handbook: Information Design for Public Places., first ed. Newyork: Princeton Architectural Press, 2009.

[2] J. D. Wineman and J. Peponis, „Constructing spatial meaning: Spatial affordances in museum design,“ Environment and Behavior, vol. 42(1), pp. 86-109, 2010.

[3] B. R. Fajen and W. H. Warren, „Behavioral dynamics of steering, obstable avoidance, and route selection,“ Journal of Experimental Psychology: Human Perception and Performance, vol. 29(2), pp. 343-362, 2003.

[4] R. G. Golledge, A. J. Ruggles, J. W. Pellegrino, and N. D. Gale, „Integrating route knowledge in an unfamiliar neighborhood: Along and across route experiments,“ Journal of Environmental Psychology, vol. 13(4), pp. 293-307, 1993.

[5] H. Seraji and A. Howard, „Behavior-based robot navigation on challenging terrain: A fuzzy logic approach,“ IEEE Transactions on Robotics and Automation, vol. 18(3), pp. 308-321, 2002.

[6] D. Helbing, I. Farkas, and T. Vicsek, „Simulating dynamical features of escape panic,“ Nature, vol. 407(6803), pp. 487-490, 2000.

Fundamental diagram as a model input – direct movement equation of pedestrian dynamics

Ekaterina Kirik, Institute of Computational Modelling RAS, Novosibirsk RUSSIAAndrey Malyshev, Institute of Computational Modelling RAS, Novosibirsk RUSSIATat`yana Vitova, Institute of Computational Modelling RAS, Novosibirsk RUSSIAAlexander Kamyshnikov, Siberian Federal University, Krasnoyarsk RUSSIA

Leadingbyadvantagesofcontinuousanddiscreteapproachestomodelpedestriandynamicswedevelopanewdiscrete-continuousmodelSIgMA.DC[5].Inthismodelpeople(particles)moveinacontinuousspaceinthissensemodeliscontinuous,butnumberofdirectionswhereparticlesmaymoveisamodelparameter(limitedandpredeterminedbyauser)inthissensemodelisdiscrete.Herewe deal with an individual approach when models give coordinates of each person.

Aspaceandaninfrastructure(obstacles)areknown.Peoplemaymovetofreespaceonly.Shapeofeachparticleisadiskwithadiameterofdi,[m],initialpositionofaparticleiisgivenbythecoordinateofthecenterofthediskxi(0)=(xi1(0),xi2(0)),i=1,...,N,N–numberofparticles.Eachparticleisassignedwithafreemovementvelocityvi0,[m/s],squareofprojection,mobilitygroup,ageofeachperson.Letassumethenearestexitasatargetpointofeachpedestrian.Intersectionsofparticles,particlesandobstaclesareforbidden.ToorientinthespaceparticlesusethestaticfloorfieldS[8].

Anideaoftheapproachproposediscomefromtheformulathatconnectspathandvelocity.It’sfinite-differenceexpressioninavectorwaygivesasanopportunitytopresentmovementequationinadirectform,whencurrentpositionoftheparticleisdeterminedasafunctionofapreviouspositionandlocalparticle’s velocity. For each time t coordinate of each particle i are given by the following formula:

xi(t)=xi(t-dt)+vi(t)ei(t)dt,i=1,...,N,(1)

wherexi(t-dt),[m,m]–coordinatesinprevioustimemoment;vi(t),[m/s]–currentvelocityoftheparticle;dt,[s]–lengthofatimestep(fixed).

Unknownvaluesin(1)arethedirectionei(t)andtheshiftvi(t)dt.Thedirectionisproposedtoberandom.Eachtimestepteachparticleimaymoveinoneoftheqpredetermineddirection,q–numberofdirections,modelparameter.Choiceofthedirectionwhereparticlesmakeanextstepisstochasticandbasedonprobabilitiesdistribution.“Right”probabilitiesvarydynamicallyandaregivenbybalancingofthreecontributions:a)themaindrivenforce(givenbydestinationpoint),b)interactionwithotherpedestrians,c)interactionwithaninfrastructure(nonmovableobstacles).ProcedureofcalculatingprobabilitiestomovetoeachofthedirectionsisadoptedfrompreviouslypresentedstochasticCellularAutomata(CA)floorfield(FF)model[2,3,4,6]andgivesthehighestprobability to direction that has most preferable conditions for movement considering other particles andobstaclesandstrategyofthepeoplemovement(theshortestpathand/ortheshortesttime).Directedmovementisgivenbyusingthestaticfloorfieldthatshowsadistancefromeachpointofthespace to the nearest exit.

Ateachtimestepcurrentvelocityvi(t)oftheparticleisdeterminedbylocaldensityinaccordancewiththefundamentaldiagram[1,7].Tobemoreexactlyweuseanalyticalexpressionsofvelocityversusdensitythatisgivenbyformulain[1].ThisformulawasderivedfrompreviouslypresentedresultsinthebookbyPredtechenskiiandMilinskii[7].Obviouslyonecanuseotherdata,(velocityvs.density)that is in table or formula form.

Intheclassofcontinuousmodelsuchapproachhasanadvantage.Usuallydifferentialequationisusedtogivemovementequationoftheparticle.TosimulatemovementofNparticlesasystemofdifferentialequationsaresolved.Itworsetobementionedthatnumericalsolutionoftheprobleminsuchstatementis“ahardnuttocrack”.Forcesthatacttotheparticlecouldnotbeadoptedfromphysicallawsdirectlyandshouldsatisfyatleastonequalitycondition:modelflowshouldcorrespondfundamentaldiagram.Todescribeforcesinsuchwayisdifficultproblem.

Intheapproachproposedweleavedapartthisstep.Themovementequationisgivenindirectform,and velocity is controlled by local density in the direct form. In an assumption that movement direction is “right” such approach should give strong coincidence of the model flow and experimental flow that isused.Sothesuccessinthepedestrianmovementmodelingusingsuchapproachistochoose“right”directionsforeachparticleineachtimestep.ApproachofchoosingdirectionsthatwascreatedforCAmodel was adopted and adapted for continuous space.

Modeldynamicswasinvestigatedforsimplegeometry,straightcorridorb=2metersinwidth.ForperiodicboundaryconditionstherewasacontrollineandwemeasuredtimeTthatN=1000peopleneededtopassthisline.TherewasobservedaperfectcoincidenceofthemodelflowJ=1000/T/b,[pers/s/m],andrealdata[1]forlowandmiddledensities.Forhighdensitiesmodelisconsiderablyslowerthanarealflow.Itseemstobeacomputationalimpactofthemodel.Thefactisthelowerfreemovementvelocity[1,1]thebettercoincidenceoftheflowsobservedforhighdensities.Experimentsonopenboundaryconditionsshowedatypicalandqualitativelythesamebehaviorthatmodelshouldreproduce,andRIMEAcollectionpresents.Experimentsonothergeometriesandfurtherdevelopmentof the model are going on.

REFERENCES

[1] Kholshevnikov, V.V., Samoshin, D.A. (2009) Evacuation and human behavior in fire, Moscow, Academy of State Fire Service, EMERCOM of Russia, 212 p. (Rus.)

[2] Kirik, E., Yurgel‘yan, T., Krouglov, D. Artificial Intelligence of Virtual People in CA FF Pedestrian Dynamics Model // LNCS, V. 6068, 2010.– pp.513-520.

[3] E. Kirik, T. Yurgel’yan, D. Krouglov. On Influencing of a Space Geometry on Dynamics of Some CA Pedestrian Movement Model // Lecture Notes in Computer Science, V. 6350, Cellular Automata, 2010. – P. 474-479.

[4] E. Kirik, T. Yurgel’yan, D. Krouglov. On Time Scaling and Validation of a Stochastic CA Pedestrian Dynamics Model // In the book “Pedestrian and Evacuation Dynamics“, editors Richard D. Peacock, Erica D. Kuligowski, Jason D. Averill, 2011, 819-822 (DOI: 10.1007/978-1-4419-9725-8_80).

[5] Kirik, E., T. Yurgel‘yan, A. Malyshev. On discrete-continuous stochastic floor field pedestrian dynamics model SIgMA.DC // In the book “Emergency evacuation of people from buildings”, 2011. – pp. 155-161.

[6] Kirik, E., Yurgel‘yan, T., Krouglov, D. (2011) On realizing the shortest time strategy in a CA FF pedestrian dynamics model // Cybernetics and Systems, vol.42:01, 1-15.

[7] Predtechenskii, V.M., Milinskii, A.I. (1978) Planing for foot traffic flow in buildings. American Publishing, New Dehli. Translation of: Proektirovanie Zhdanii s Uchetom organizatsii Dvizheniya Lyudskikh potokov, Stroiizdat Publisher, Moscow, 1969.

[8] Schadschneider, A., A. Seyfried. Validation of CA models of pedestrian dynamics with fundamental diagrams // Cybernetics and Systems, Vol. 40 (5), 2009, 367–389.

Experimental study and modelisation of pedestrian space occupation and motion pattern in a real world environment

Francesco Zanlungo, Advanced Telecommunication Research Institute, Kyoto JAPANYoshihiro Chigodo, Osaka University, Osaka JAPANTestushi Ikeda, Advanced Telecommunication Research Institute, Kyoto JAPANTakayuki Kanda, Advanced Telecommunication Research Institute, Kyoto JAPAN

Thepurposeofthisworkistostudythebehaviourofpedestriansinarealworldenvironmentwithaverysimplegeometry.Ideally,wewouldliketostudyalong,uniformcorridorwhichisusedjustasatransitionplace,i.e.ithasnolateralexitsorentries,shopsorotherpointsofinterest.Furthermore,thecorridorshouldnotpresentanynarrowingorwidening,noranyarchitecturalelementsorequipmentsthatcaninfluencethewaypedestriansusethespace,andallthepedestriansintheareashouldbehaveascommuters(nogatheringsandthelike).Suchanenvironmentisexpectedtobesymmetricalalongthedirectionofthecorridor,andwewanttostudyhowthepedestriandensity,velocityandfluxchangeasafunctionofthedistancefromoneofthecorridor‘swalls(i.e.inthedirectionorthogonaltothecorridor axis).

AsanapproximationtothisidealenvironmentwedecidedtostudyanundergroundareainUmeda(Osaka)wheresomecorridorsconnectashoppingareawitharailwaystation.Thesecorridorsarequiteuniform,withoutanyshop,andusedalmostonlybypeopletransitingbetweenthestationandtheshoppingarea,andthustheirstructureisquitesimilartotheidealone.Thepedestrianpositionswererecordedintwoworkingdayafternoonsusing2Dlaserrangefinders,atechnologythatallowsforautomaticdetectionwithanerroroforder5centimetres[1].Thistechnologysuffersofclutteringproblemswhenthedensitygetsveryhighbutworksverywellatthedensitiesoccurringinthesecorridors.

Wedividetheenvironmentin2dimensionalsquarecellsoflinearsize250mmanddefineoneachofthem5observables,namely:thepedestriandensityrhoasthenumberofpedestriansrecordedinthecelldividedbythenumberofobservationtimeinstantsandcellarea;averagevectorialvelocity;averagespeed;currentjdefinedastheproductofrhoandaveragevectorialvelocity;andflowphidefinedastheproductofrhoandaveragespeed.Analysingthesequantitiesweareabletospotandcliptheareasoftheenvironmentthatbettersatisfyourrequirementsforauniformcorridor,andthenfurtherfilterthedatatoexcludefromourstatisticalanalysisthosepedestriansthatarenotmovingalongthecorridor‘saxis(thisfilteringprocessisvalidprovidedthattheamountofexcludeddataisnegligiblewithrespecttothetotalamountofdataintheclippedarea).Thefiltereddataarethenintegratedonthe(symmetrical)corridordirectioninordertoobtain1Ddistributions.

Wefindthatthejdistributionhastwoextremasincethefluxofpedestriansinthecorridorsplitsintwoaccordingtothewalkingdirection(peoplewalkontheleftsideofthecorridoraccordingtotheJapanesetrafficconvention),withthetwofluxesassumingclearlyseparatedmaxima.Nevertheless,atleastinthestudieddensityregime,thetwofluxesoverlapresultinginanalmostconstantphidistributionbetweenthetwojmaxima.Thephidistributiondropstozeroclosetothewalls.The

rhodistributionisverysimilartothephione,suggestinganalmostflatdistributionforthespeed.Accordingly,wefindthatthespeeddistributionisaconvexfunctionassumingamaximuminthecentreofthecorridorandminimaclosetothewalls,butthedifferencebetweenthemaximumandtheminimaisnobiggerthan10percent.Interestingly,evenifwestudythespeeddistributionfortheindividualfluxes,wefindthemaximumtobeatthecentreofthecorridor,i.e.ontherightofthethejmaximum,suggestingatendencytoovercomeontheright.

Weproposeatheoreticalmodeltoexplainthedistributionofpedestriansinagivenflux.Denotingwithxthedistanceofthepedestrianfromoneofthewallsweintroducean`ènergyfunction‘‘(actuallyacomfortfunction)E(x)thathasdivergingtermsonthewallsandisquadraticinthedistancefromagiven point c (the point where the pedestrian attains minimum “energy” i.e. maximum comfort).Assuming,followingtheprinciplesofstatisticalmechanics,thattheprobabilitydistributionofpedestriansisgivenbyaBoltzmannfactor,p(x)=exp(-E(x)),wefindresultsingoodagreementwiththerhodistributioninourdata,providedthatthequadratictermintheenergyislimitedtoaboundedvaluetoexplainthefiniteprobabilityofwalkingonthe``wrong‘‘sideofthecorridor.Calibrationonrealdataquantitativelyshowsthatcisalwaysclearlylocatedintheleftsideofthecorridor.Theenergyfunctioncouldbeextendedinthe(x,v)spacetoaccountalsoforthevelocitydistribution,whichcandescribedquitewellbyaGaussianfunction(evenifthemodeisalwaysslightlyhigherthanthemean;thedistributioncannotobviouslybedescribedasMaxwell-Boltzmann,sincevelocityhasaclearlydefineddirection);ifanalysedfordifferentpositionsinthecorridor,themeanvalueoftheGaussianvelocitydistributionassumesamaximumvalueatthecentreofthecorridor,andis,foreachflux,biasedtowardtheright,suggestingonceagainatendencytoovertakeontheright.

Itwassuggested[2]thatthetendencytowalkonagivensideofacorridorcanberelatedtoa(culturallydependent)biasincollisionavoiding.Withsimulationsbasedondifferentmicroscopicpedestrianmodels[3,4]weverifyifthisassumptionisinagreementwithourempiricalfindingsandstatisticalmodel.Furthermoreweinvestigatewhichkindofmodificationshavetobeintroducedinthemodelstoaccountfortheobservedvelocitydistribution,andweinvestigatewhicharetheeffectsofthesemodificationsonsimulationofsystemsatdifferentdensities.

REFERENCES

[1] D.F. Glas, T. Miyashita, H. Ishiguro, N. Hagita, “Laser-Based Tracking of Human Position and Orientation Using Parametric Shape Modeling” , Advanced Robotics, 23 (4), 405-428 (2009)

[2] M. Moussaid, D. Helbing, S. Garnier, A. Johansson, M. Combe and G. Theraulaz, “Experimental study of the behavioural mechanisms underlying self-organizaion in humancrowds”, Proceedings of the Royal Society B: Biological Sciences, 276, 1668, 2755-2762, (2009)

[3] A. Johansson, D. Helbing, P. S. Shukla (2007), “Specification of the social force pedestrian model by evolutionary adjustment to video tracking data”, Advances in Complex Systems, 10, 271-288.

[4] F. Zanlungo, T. Ikeda and T. Kanda, “Social Force Model with explicit collision prediction”, Europhysics Letters, 93, 68005 (2011).

Psychological aspects of german signal words in evacuation warnings

Laura Künzer, Friedrich-Schiller-Universität Jena, Jena GERMANYGesine Hofinger, Friedrich-Schiller-Universität Jena, Jena GERMANYTina Zink, Universität Regensburg, Regensburg GERMANY

Warningsplayanimportantroleinevacuation.Theyaimatavoidingorreducinginjuryandlossoflife(Wogalter&Laughery,2006).Therecipientsmustbeabletoactaccordingtothewarningwhichimpliesunderstandingthemeaning,knowingwhattodo(leavethesite)anddoingit.Butoften,themeaningofawarningisnotclearoritcannotbeproperlyunderstood.Thisproblemofpoorlydesignedwarningshasbeenaddressedforsometime.AsalreadyreportedinSime(1995),duringanunannouncedevacuationexerciseinanundergroundstationinGreatBritain,thewarninggivenwasveryimpreciseanddidnotcontainspecificinstructions.Passengersgatheredinformationinsteadofleavingthestationimmediately.Incaseoflackofinformationorunclearmeaning,peoplewillspendtimeforcollectinginformationinsteadofleavingthesite(Tubbs&Meacham,2007).Thus,thetimebeforeevacuationstartscouldbeshortenedsignificantlybyearlyandpreciselygiveninstructionsincluded in warnings.

Itisessentialforsafetytoprovideadequateinformationaboutthedangerathand.Insomesituations,situationalcuessuchassmoke,heatornoiseserveaswarningsignals.Yet,peoplethatarepotentiallyaffectedmightdetectthemtoolateinordertorecedesafely.Inmanycases,explicitwarningsignalsareeventheonlysourceofinformationforthepersonsconcerned(Edworthy&Adams,1996).

Warningsthereforehaveanalertingfunctionandaninformingfunctionregardingthesourceandconsequencesofdangeraswellasthenecessaryaction(Wogalter&Laughery,2006).Animportantpartofvisualandverbalevacuationwarningsaresignalwordssuchas“attention”,“warning”,“notice”.Theyalertattentionwhiletransportinginformationaswell(Edworthy&Adams1996;Hellier,Aldrich,Wright,Daunt&Edworthy,2007).Butnoteverysignalwordusedinnaturallanguageisappropriateforeverymessage.Theresponsetothewarningwillpartlydependontheconnotationsorsemanticfieldofthewordsused.Explicitnessisimportant,aswellasperceivedurgency.Whendesigningawarning,psycho-acousticurgencyofthesignalwordusedshouldmatchthecontextualurgency(urgencymapping,e.g.Edworthyet.al,2003).Inthatway,recipientsintuitivelyunderstandthedangerathand,reactmorequicklyandshowmorecompliancewiththemessage(Laugheryetal.,1993).

ForEnglishsignalwords,thesemanticfieldhasbeenresearchedforatleasttwodecades(e.g.,Edworthyetal.,2003Hellieretal.,2007;Wogalter&Laughery,2006).Relevantdimensionsareperceivedhazard,arousalstrength,intendedcarefulness,andperceivedurgency.Forexample,independentofmodality,“note”,“notice”,“attention”seemtoconveylessdangerandurgencythan“danger”or“warning”(Edworthyetal.,2003).ForGermansignalwords,nostudieswerefound.So,forthedesignofeffectiveevacuationwarnings(loudspeakerannouncements/publicannouncements),astudyonsignalwordsusingtwoexperimentswasconducted.TheaimofthestudyisthemappingofGermansignalwords

onthedimensionsofperceivedurgencyandexplicitness.Infirstteststheauthorshadfoundthatthemeaningofthesignalwordsvariedwithcontext.Inconsequence,aspecificcontextofloudspeakerannouncement in underground stations was chosen because the results of the study are to be used in theresearchprojectOrGaMIRPLUS.Thisproject,fundedbytheGermanMinistryofEducationandResearch,dealswithaspectsofhumanfactorsinevacuationofundergroundtransportationsystems.

Intwoexperiments,theperceivedurgencyandexplicitnessofvisuallypresentedsignalwordswasanalysed.WechosesignalwordswhicharesimilartothosealreadyusedinEnglishstudiesandstandardsandwhichcanbefoundinGermanwarningsaswell(„Achtung“-attention,„Gefahr“-danger,„Hinweis“-notice,„Vorsicht“–beware/becareful,and„Warnung“–warning).InExperiment1(n=51),thesignalwordswerepresentedaswrittentext.Subjectsansweredeightquestionsusingstandardizedratingsscalesregardingthefivesignalwords.Subjectswereinstructedtoimaginetheywerewaitingforthetrainwhenanannouncementstarted.InExperiment2(n=36,noneofthemparticipatedinexperiment1),thesignalwordswerepresentedinspokenlanguage(recordingsbyaprofessionalfemaleannouncer).Subjectslistenedtothesignalwordsandtoannouncementsbeginningwithsignalwordsandthenansweredthesamequestionsusingratings.

Weexpectedtofindaclearrankingorderbetweenfivesignalwordswhiletherankingorderforurgencyshouldbethesameasforexplicitness.ForExperiment1,datashowedthat“Gefahr”(danger)isperceivedasmosturgentandmostexplicitwhile“Hinweis(notice)isleasturgentandleastexplicit.Butfortheotherthreesignalwords,noclearordercouldbefound.Rankingordersforexplicitnessandurgencywerethesame.Resultsforexperiment2arestillbeinganalysed.Thesefindingshaveadirectrelevanceforevacuationwarningsbecause“Achtung”(attention),“Warnung”(warning)and“Vorsicht”(beware/becareful)aremostoftenusedinwarnings,alsoinundergroundloudspeakerannouncements.Thoseresponsibleforevacuationrelyonsignalwordsthatconveyquitevaryingmeanings to the listeners.

REFERENCES

Edworthy, J. and Adams, A. (1996). Warning design: A research prospective. Oxford: Taylor & Francis.

Edworthy, J., Hellier, E., Walters, K., Clift-Mathews, W. & Crowther, M. (2003). Acoustic, semantic and phonetic influences in spoken warning signal words. Applied cognitive psychology, 17, 915-933.

Hellier, E., Aldrich, K., Wright, D. B., Daunt, D. and Edworthy, J. (2007). A multi dimensional analysis of warning signal words. Journal of risk research, 10 (3), 323-338.

Parsons, S. O., Seminara, J. L. and Wogalter, M. S. (1999). A summary of warning signal research. Ergonomics in design, 7 (1), 21-31.

Sime, J. D. (1995). Crowd psychology and engineering. Safety Science, 21(1), 1-14.

Tubbs, J. S. and Meacham, B. J. (2007). Egress design solutions: A guide to evacuation and crowd management planning. Hoboken: John Wiley and Sons.

Wogalter, M. S. and Laughery, K. R. (2006). Warnings and hazard communications. G. Salvendy (Hrsg.), Handbook of human factors and ergonomics (3rd ed., S. 889-911). Hoboken: John Wiley and Sons.

Effect of guidance information and human relations among agents on crowd evacuation behavior

Tomoichi Takahashi, Meiji University, Tokyo JAPANMasaru Okaya, Meiji University, Tokyo JAPAN

INTRODUCTIONEvacuationofpeopleintimesofemergencyordisasterisacomplextask.Itisdifficulttoconductphysicalexperimentsthatinvolvemanyhumansandrealenvironments,ortomakeeffectiveplansforpredictablesituationsbasedonthedataofpastdisasters.Disastersimulationsystemshavebeenemployedtomakethepreventionplansforthedisastersandtochecktheplans.Insimulationsystems,evacueeshavebeentreatedequally,orthedifferenceinageandsexoftheevacueeshavebeeninvolvedastheparametersofsimulations[1].

Anagentbasedcrowdevacuationsimulationisproposedinthispaper,tosimulateevacuationbehaviorsthatcontainfollowingphenomena.Thephenomenawerereportedtooccurinevacuations[2].1.Rescueworkerswhoenterintoabuildingmoveintheoppositedirectionofevacueeswhogooutofbuildings.Themovementcausesinteractionsintheevacuations.Family-mindedhumanbehaviorssuchasparentsseekfortheirchildren,mayalsocausesimilarinteractionincrowdevacuation.2.Evacuationguidanceleadspeopletoevacuatesafelyandefficiently.Whilesomepeoplestarttoevacuateimmediatelywhentheyheartheevacuationguidance,othersmaycontinuedoingworksiftheguidanceisnotclearlyannounced.Theothermaynotheartheguidanceattheworstcase.

FEATURESOFOURSYSTEMANDEVACUATIONSCENARIOSWeproposeasystemthatsimulatestheevacuationbehaviorofcrowdwiththephenomenabyconsidering human relation among agents and guidance announcement at emergencies. Features of the system are followings:1.Inemergencies,humanbehavesdifferentlythanusually.Thebehaviorsareaffectedbypeople’smentalcondition.BDI(Belief-Desire-Intention)modelisemployedtopresenthowagentsselecttheactionsaccordingtothesituationsatsense-reason-actcycle.2.Parentstakecareoftheirchildren.Theiraltruismforcethatcomesfromhis/herownhumanrelationsisaddedtoasocialforceinHelbing’smodel[3].Itworkswhenaparentwaitstillhisorherchildcatchesup,forexample.Thismaybecomeblockadestootherswhoharrytorefuges.3.Indisasters,anevacuationroutemayberenderedimpassablebyrubble.Theevacueesexchangeinformationonthedynamicallychangingsituationsamongthem.Twotypesofinformationpropagationsareintroduced;communicationbetweenasecurityofficerandanevacuee,andcommunication among evacuees.

Following three scenarios are presented to show our features that traditional evacuation systems have not had.

Scenario1.Pairsofparent-childevacuationfromcampus:Aneventisheldatacampusthatconsistsoftwobuildings.Manypairsofparent-childparticipatetheevent,andtheyaredividedintotwogroups.Whendisastersoccur,theyaresupposedtoevacuatefrombuildingstoanearbyrefuge.Followingthreecasesaresimulatedandresultsshowdifferentevacuationbehaviors although there is the same number of agents in each building.a)Parentsandtheirchildrenareinthesamebuilding.b)Agentsarerandomlylocatedsothatsomeparentsandtheirchildrenareindifferentbuildings.c)Forallparent-childpairs,parentsandtheirchildrenareindifferentbuildings.

Scenario2.Evacuationwithguidancefromsecurityoffice:Inanemergency,peoplegotothenearestrefuge.Whenfireoccursnearoneofrefuges,asecurityagentannouncesthatitissafertogootherrefugesthantherefugenearthefire.Someagentsfollowtheguidance,whileotheragentsdonot.Insomesimulations,therearephenomenathatsomeagentsthatdon’tfollowtheguidanceareinvolvedinthemovementsofanumberofpeople.Theirmotionsthatgotothedifferentrefugeisoppositetotheothers’movementsandcausetrafficjams.

Scenario3.Evacuationguidanceandcommunicationamongcivilians:Duringtheevacuationfromabuilding,occupantsexchangeinformationondisasters.Accordingtotheinformation,someoccupantschangetheiractionsaccordingtotheirknowledgeonemergencystairsinthebuilding.Thechangesofactionaffecttheevacuationtimefromthebuilding.Simulationresultsshowbetterqualityofguidancemakesevacuationtimeshorter.

SUMMARYANDDISCUSSIONRecently,peoplehavebeenkeentoassessthesafetyofsocietyafterseveraldisasters,andanalysisofevacuationbehaviorhasreceivedincreasedattention.Theagent-basedsimulationprovidesaplatformfor computing individual and collective behaviors in crowds.

Wepresenttwoideas;anagentbehaviormodelpresentedwithBDIandinformationpropagationthroughcommunication,togeneratethephenomenaobservedincrowdevacuations.TheBDImodelpresentsthehumanrelationsamongagentsandpatternsofitsownbehaviors.Thecommunicationenables to represent how much agent trust the information that they get through communication amongagents.Thesimulationresultsoftheevacuationscenariosrevealthefollowings:1)Propertyofhumanrelationsandinformationintheguidanceaffectevacuationbehaviors.Thebehaviorsaredifferentfromonesoftherestpeople.2)Thebehaviorsofsomeagentsaffecttheevacuationbehaviorsinall.Itcausescongestionandittakesmore time to evacuate as in real life.Theseresultsdemonstratethatourmodelprovidesaneffectivesimulatingmethodofcrowdbehaviorinemergencysituations,andthatchecktheevacuationplansandtheannouncementofguidance.

REFERENCES

[1] G. Santos and B. E. Aguirre: A Critical Review of Emergency Evacuation Simulation Models, NIST Workshop on Building Occupant Movement during Fire Emergencies, 25-50, June 9-10, 2004.

[2] J.D.Averill: NIST ncstar 1-7: Occupant behavior, egress, and emergency communication, Sep. 2005.

[3] D.Helbing, et. al: Simulating dynamical features of escape panic, NATURE, 487-490, Sep. 2000.

Using the social force model to represent the behavior of pedestrians at chaotic intersections of developing countries: the case of Peru

Felix Cabrera, Pontificia Universidad Catolica del Peru, Lima PERUJuan Carlos Dextre, Pontificia Universidad Catolica del Peru, Lima PERU

INTRODUCTIONInthecitiesofPeruanimportantsetofintersectionspresentproblemsofcongestion,pollutionandmainlyaccidents.Thestudyby[1]indicatesthatamong178countriesconsidered,Peruisthefirstinpedestriandeathsfromroadaccidents(78%).Manyoftheseproblemsaregeneratedbecausethetraditionalmanagementsystemgivesprioritytotrafficinsteadofvulnerableusers(pedestrians,thedisabled,cyclists,etc.).Peruvianpoliticiansprivilegetomotorvehiclesandthinkthatanyimprovementfor the other user groups would imply loss of capacity on the roads and more travel time and delays for vehicles.

DuetomismanagementandthecharacteristicsofPeruvianspedestriansanddrivers,thereareseveralconflictsbetweenpedestrians,cyclistsandvehiclesthatcannotberepresentedbytraditionaltrafficmodelingtools.StudiesinsomecitiesofPeru[2]indicatethatanalyticalmodelswouldnotbeadequatetoanalyzetrafficsituationsclosetocongestionorwhereaggressivedrivingandmultiplepointsofconflict between the modes are present.

InordertoassessaccuratelythebenefitsofapplyingmitigationactionsisnecessarytousetogetheraroadsafetyprocedurewithtrafficandpedestrianmicrosimulationmodelthatcanbeableofrepresentingthebehaviorofPeruvianpedestriansanditsinteractionwithothertransportmodes[3]Traditionally,thebehaviorofpedestrianshasbeenmodeledasiftheyweregases[4-5]andfluid[6,7,8]andpedestrianareashavebeenanalyzedinnormaltrafficconditionsandevacuation.InthisparticularcasePedestriansmovegovernedmainlylookingforthesafestandfastestroute,whichisnotnecessarilythroughthecrosswalk.Thereforeweanalyzeasituationinnormaltrafficconditions,includingrandomroutes and where within the space of the intersection (roads) are conflicts between pedestrians andvehicles.Forthisreasonitisthoughtthatthesocialforcemodel[9]mayrepresentthischaoticbehavior.

OBJECTIVESTheaimofthisstudyistoinvestigatewhetherthesocialforcemodelwouldbeabletorepresentthechaotic movement of pedestrians at an intersection and the conflicts that occur with other modes of transport.Inaddition,itseekstodeterminewhetherthemitigationmeasureswouldadverselyaffectvehiculartrafficasPeruvianpoliticiansthink.

METHODOLOGYThemethodologywastoevaluateasignalizedintersectioninthecityofChiclayo(northofPeru)withahighpercentageofvulnerableroadusers(especiallypedestrianswhotakeanyrouteandmoveamong

vehicles).WeusedthetrafficmicrosimulationsoftwareVISSIManditspedestrianmodule(basedonthe social forcemodel).

Thedataformodelconstruction,calibrationandvalidationwerecollectedduringtwoworkingdaysfrom11:00amto2:00pm,bymanualscountingsandfilms,andthepeakhourofeachdaywasmodeled.Thetrafficcalibrationandvalidationwasperformedwithtrafficvolumesandqueuelengthsattheentrances.Forthis,theparametersofthecarfollowingWiedemannmodelwereadjusted.Also,thevolumesofpedestriansandtheirroutesfromthesimulationmatchedthefieldmeasurements.Twentyfiverunswereperformedforeachsetofparametersuntiltherewasnoevidenceofdifferencebetweenthefieldmeasuredvaluesandtheaverageonesobtainedwiththesimulation.Theconfidencelevelwas95%.

Thecurrentscenario,calibratedandvalidated,thatshowsconflictsamongpedestrians,vehicles,peopleinwheelchairanddisabled,wascomparedwiththealternativescenario(withmitigationmeasures).Amongthemitigationmeasureswecanfindchangesinthenumberoftrafficlightphases,sidewalksandcrosswalkswider,alsotheconstructionoframpsthatallowthemovementofpeopleinwheelchairs,etc.Theproposalsfocusedonprovidingfurtherfacilitiesforvulnerableroadusers,improvedthepublicspace,andchangecontroldevicesintersectionstoimprovesafetyingeneral.

RESULTSTheresultsindicatethatmitigationmeasureswouldallowimprovepedestrian’ssafetyandtheirspeed(25%higher).However,notonlypedestrianarefavoredbutalsotrafficcirculation.Thereisa35%reductionindelaysandincreasedaveragevelocityof65%.

CONCLUSIONSThesocialforcemodelcouldbeusedtoreplicatethebehaviorofpedestriansandtrafficinteractionatintersectionsinPeru.ThevideogeneratedfromtheVISSIMsoftwareshowsthechaoticintersectionanalyzedanditssimilaritywiththereality.Inaddition,theresultswouldindicatetoauthoritiesthatthemanagement measures implemented to improve chaotic environments favor to all users and reduce the possibility of accidents with pedestrians.

REFERENCES

[1] WHO (2009). Global status report on road safety. It‘s time to take action. World Health Organization. Geneva.

[2] Cabrera (2010). ¿Can deterministic models and micro simulation analyze properly the signalized intersections in Lima

[3] Cabrera. F, Dextre, J. (2010). Use of check lists and microsimulation to improve road safety of vulnerable users.

[4] Helbing (1992). Complex Syst. 6, 391

[5] Helbing (1993). Stochastische Methoden, nitchtlineare Dynamik und quantitative Modelle sozialer Prozesse

[6] Helbing (1990). Diplom thesis, Georg-August University, Gottingen, Germany.

[7] Henderson (1971). Nature 229, 381

[8] Henderson (1974). Transp. Res. 8, 509

[9] Helbing, D, Molnar P. (1995). Social Force model or pedestrian dynamics

Keynote


Marjolein de Jong, Hasselt University, Hasselt BELGIUM The characteristics and needs of pedestrians with mobility impairments: How to move around comfortably and safely with a reduced ability to walk, see, hear, feel or process information

Mobility is one of the preconditions for being able to participate in social life: individuals performactivitiesbecauseofeconomic,social,recreationalandotherpersonalreasons.Mobilityconstraintsmayleadtoadecreasedparticipation(WorldHealthOrganization,2001).Makingthemobilitylandscapemoreinclusiveisnotaneasytaskatall:thewholetravelchainhastobedesignedandorganizedinsuchawaythatthespecificneedsofseveralgroupsaretakenintoaccount.

Themobilityneedsofpedestrianswithmobilityimpairmentscanbegroupedaccordingtothreespecificbarriercontexts:1)travellerschallengedbytemporaryconstraints,2)activetravellershappytoacceptsomesupport,and3)travellerswithconstantconstraints.Theaccessibilityofanenvironmentisnotonlydeterminedbythebarriercontextoftheperson,butalsobytheenvironmentandtheactivityundertaken.Dependingontheseconstraints,peoplewithreducedabilitiestowalk,see,hear,feelorprocessinformationcanmovearoundmoreorless independently.

Oral presentations


A.7 Michael Schultz, Technische Universität Dresden, Dresden GERMANY Stochastic transition model for pedestrian dynamics

A.8 Taku Fujiyama, University College London, London UNITED KINGDOM The effects of the design factors of the train-platform interface on pedestrian movements

A.9 Dietmar Bauer, AIT Austrian Institute of Technology, Wien AUSTRIA Including route choice models into pedestrian movement simulation models

B.7 Dirk Hartmann, Siemens AG, München GERMANY Dynamic medium scale navigation using dynamic floor fields

B.8 Paul Townsend, Crowd Dynamics International, Knutsford UNITED KINGDOM Pedestrian gap acceptance in micro-simulation modelling

B.9 Frank Huth, Technische Universität Berlin, Berlin GERMANY A macroscopic multiple species pedestrian flow model based on heuristics imple mented with finite volumes

C.7 Mohcine Chraibi, Forschungszentrum Jülich GmbH, Jülich GERMANY A simplified force model and enhanced steering for a quantitative description of pedestrian dynamics

C.8 Mario Campanella, Delft University of Technology, Delft NETHERLANDS Quantitative and qualitative validation for general use of pedestrian models

C.9 Naveesh Reddy, PM Dimensions Pvt. Ltd., Gandhinagar INDIA FDS+Evac model validation for seated row arrangements

Stochastic transition model for pedestrian dynamics

Michael Schultz, Technische Universität Dresden, Dresden GERMANY

Weproposeacalibratedtwo-dimensionalcellularautomatonmodeltosimulatepedestrianmotionbehavior.Itisav-max=4(3)modelwithexclusionstatisticsandrandomshuffleddynamics.Theunderlyingregulargridstructureresultsinadirection-dependentbehavior,whichhasinparticularnotbeenconsideredwithinpreviousapproaches.Weefficientlycompensatethesegrid-causeddeficiencieson model level.

Thedifferentmodelapproachesformicroscopicpersondynamicsarebasedontheparticulardisciplineanalogies,rangingfromhydro-dynamicmodelstoartificialintelligenceandmulti-agentsystems.Thecomplexdynamichumanbehaviorisinducedbyindividualdecisions,whichareclassifiedtobeofshort-range(operational)andlong-rangetype(strategic/tactical).Theself-organizationofpersonsis a further essential characteristic of human behavior. In contrast to the social force model or the discrete choice model the developed motion model is based on a stochastic approach to handle the unpredictablebehaviorbyindividualpathdeviations.Thestochasticmotionmodelisanappropriateandfastmethodforanalysisthedynamicpedestrianbehavior.However,toderivevalidresultsseveralsimulationruns(>100)havetobeperformed.Thefocusconcentratesontheevaluationofapplicationorientedsimulationscenariosinsteadofthecharacteristicsofindividualinteractionsorspecificpedestrian trajectories.

Thepresentedmotionmodelisbasedonastochasticapproach,whichiscomparabletoacommoncellularautomaton.Itutilizesaregulargridstructure.Incontrasttothecellularautomaton,thenew model is developed on the basis of a fundamental paradigm shift: instead of changing the cell statusdependingonthestatusofitssurroundingcells(neighbors),theagentisabletomoveovertheregularlatticeandtoenterthosecells,whicharenotoccupiedbyotheragentsorobstacles(e.g.walls).Todescribethemotionbehaviorofanagent,themotionvectorisseparatedintoadesiredmotiondirectionandatransversaldeviation.Usingthespatiallydiscretegridstructureanddefiningthreetransitionstates(forward|stop|backwardorleft|ontrack|right)thenormalizedtransitionprobability(p)intothesestatescanbegenerallydefinedandsolved.

Finally,themotioncomponentsarecombinedtoa3x3transitionmatrix.Infact,thetransitionmatrixpossessesatwo-dimensionalcharacteristic,butitonlydefinesanone-dimensionaltransitionconsideringatransversaldeviation(1.5-dimensional).Toallowforathree-dimensionalagentmotionbehavior,twoindependentmotiondirectionsareneeded.Basedonadevelopedhorizontaltransitionmatrixadiagonaltransitionmatrixisderivedbyre-indexingthehorizontalmatrix.Themotiondirectionisintegratedintothestochasticmodelbysuperposethesematrices.Therotationofthissuperposedmatrix(4-foldsymmetry)allowsfordeterminingtheentirespectrumofthemotion

direction.Theunderlyingregulargridstructureresultsinadirectiondependentbehavior(e.g.enteringdiagonalcellsimplieswalkingalongerwayincomparisontohorizontallylocatedcells).

Duetotheutilizationofaregulargridstructure,thetransitionmatrixdoesnotfulfillthecriteriaofindependentagentmotionbehavior.So,thespeedandthevarianceoftheagentdependontheagentmotiondirection.Iftheagententersdiagonalcellshiswalkingdistanceislonger(approx.41%)incomparisontotheuseofhorizontallylocatedcells.Thismodelconstraintisequivalenttoasignificanthighermotionspeeddependingonthedirectionofmotion.Algorithmstocompensatethisgrid-basedspeedeffectcanbefoundatthescientificcommunityaswell.Adetailedmodelanalysispointsoutthattheexpectedvalueofthetransitionmatrix,definedbycellbasedtransitionprobabilityandtherelativelocationdiffersfromexpectedvaluewhichisspecifiedinthemotionmodel.

Theangledependedvarianceofmotionimpliestwomajorissues.So,thebasicmodelshowsanimmanentmotiondeviation,whichisnotconsideredinpreviousequationsandthedifferentvariancesleadtodifferentavoidingbehavior.Usingthesamescenariobutwithadifferentmotionangle,theagentgetsdriftingprobabilitiestopassblockedcells.Toensurehomogeneousvariancedistribution,anappropriatecompensationonmodellevelisneeded.Theexpectedvalueofthematrixdependsonspeedandstandarddeviation,whereastheparametersaredirectlycoupled.Foreachparametersetaspecificcharacteristicovertheanglehastobecalculated.

Further model investigations point out that the model has to be extended to reproduce the representativeshapeofthefundamentaldiagram.Usingthetransitionmatrix,anagentisabletoreacttothestatusoftheadjacentcells(empty,occupied).Thefundamentaldiagramindicatesaninteractionrangeofabout1.3m,becausethespeedofanagentstartstodecreaseifthedensityrelationsreachesalevelof10%maximumdensity(consideringanagentwithadimensionof0.4x0.4mandamaximumdensityof6.25pedestrian/m^2).Iftheagentmovesthree/fourstepsatonce,hewillbeabletointeractwith distant agent and the developed model is found to reproduce the characteristic shape of the fundamentaldiagram.Thereforethemotionmodelhastoprovidethefollowingagentproperties:-Alwaysmove,nowaiting(occupiedcellsincreasetransitionprobabilityoftheothermatrixcells).-Movefourstepsatonceanddecreasethestepsdependingonagentdensity,atleastinthecaseof60%ofthemaximumdensitythenumberofstepsshouldbereducedfromfourtothreetofitthefundamental diagram.-Agentleavesatrace,ateachtimestepallenteredcellswilltemporarilyblocked.

Thestochasticmodelmeetsallcriteriaforascientificallyreliablemotionmodel.Itexhibitstheabsenceofsignificantmodel-causedlimitationsandreproducesallcommonself-organizingeffects(e.g.rowformationoroscillation).Besidestheoperationalmotiondefinitionbythestochastictransitionmatrix,strategic/tacticalmotioncomponentsaretakenintoaccountaswell.Thestochasticmodelallowsforthereactionoftheagenttoobjects/agentsatimmediatevicinity.Itadditionalprovidesthecapabilityofconsideringdistantconstellationofagents(jam)andpotentiallyblockedbottlenecks.

SUMMARYANDOUTLOOKModelspecificparametercorrectionsensurethatthemotionvectorisequaltotheexpectedvalueofthecorrespondingtransitionmatrix.Thisissuehasinparticularnotbeenconsideredwithinpreviousapproaches.Thecalibratedmotionmodelisthusthefirstapproach,whichallowsforaspecificstochasticdescriptionofagentmovementswithoutmodelrestrictions.ThepassengerrelatedevaluationsandsimulationsofdispatchprocessesatDresdenAirportexemplarilyshowthat the developed stochastic motion model is able to reproduce the behavior of passengers in an

appropriateway.Thereforethedevelopedmotionmodelisimplementedinacorrespondingapplicationenvironmentwhichallowsforavariousapplication,e.g.investigationsregardingtogroupdynamicbehavior or route planning in the airport terminal focused on normal operations and emergency cases.

The effects of the design factors of the train-platform interface on pedestrian movements

Taku Fujiyama, University College London, London UNITED KINGDOMNick Tyler, University College London, London UNITED KINGDOM

BACKGROUNDANDOBJECTIVEAspedestriansimulationmodelsareusednotonlyinevaluationsimulationbutalsointheplanninganddesignofpedestrianinfrastructure,itisnecessarytoobtaindataofhowpedestriansinteractwithexternal environments (e.g. how pedestrians interact with a step on the pavement). In the transport engineering,pedestrianmodelsareoftenusedtoassessnewdesignsofstationplatformsandrollingstock.Onemajorapplicationareaistraindwellingatthestationonthebusycommuterlinebecauseevenatinyreductioninthedwelltimecanleadtoasignificantimprovementinthelinecapacityinthepeaktimeandthereforeitisessentialtoassessvariousdesignsintheplanningstageandchoosethebestoption.Inthiscase,thekeyelementisthedesignsofthetrain-platforminterface(e.g.doorwidth,sizeofthegapbetweenthetrainandtheplatform).However,therehasbeenalackofempiricaldataonhowthedesignfactorsofthetrain-platforminterfaceaffectpedestrianmovements.Althoughtherehavebeensomeobservationalstudies(e.g.Atkins,2004;Harris,2006;Wiggenraad,2001),suchstudieshavenotvarieddesignfactorsduetothenatureoftheobservationalstudy.Anexperimentalstudywouldbeusefulinordertosystematicallyinvestigatehoweachdesignfactor(e.g.doorwidth)affectspedestrianmovements.Therefore,weconductedaseriesoflaboratoryexperimentstoexaminetheeffectsofeachdesignfactorofthetrain-platforminterfaceonpedestrianmovements,particularlyonthe pedestrian flow rate.

METHODOLOGYheexperimentswerecarriedoutatPedestrianAccessibilityMovementandEnvironmentLaboratory(PAMELA)ofUniversityCollegeLondon,whereamock-uptrainandaplatformweresetup.Themock-uptrainwasahalfsizeofarealtraincarriage.Theexperimentlastedfivedays,andaround120participantsjoinedeachexperimentday.Intotal,224runswereconducted.Theexperimentwasrepresentation of a crowded commuting situation.

Intheexperiment,thefollowingdesignfactorswereinvestigated:• thedoorwidth,• thesizeoftheseatset-back(thedistancebetweenthedooredgeandtheadjacentseat)• thesizeoftheverticalgapbetweenthetrainandtheplatform,and• thesizeofthehorizontalgapbetweenthetrainandtheplatform.

Ineachrun,50passengersmademovementswhereasothersstayedinthetrainorontheplatform.Variouspedestrianmovementpatternsweretested,namely:alighting-dominant(45pedestriansalightedwhereas5pedestriansboarded),boarding-dominant(45boarded,5alighted)andeven-alighting-boarding(25alighted,25boarded).

RESULTSANDDISCUSSIONAswearestillanalysingthedata,thissectionshowssomepreliminaryresultsoftheexperiments.Wewillfinishthedataanalysisbeforethedeadlineofthepapersubmissionandinthepaperwewillpublish the results of our analysis.

First,alargerdoorwayincreasedthepedestrianflowrate.Performanceofthealighting-mainrunsespeciallyimprovedwhenthedoorwaywasincreasedfrom1500mmto1800mm.Thismayberelatedtothehumanbodysize.Theshoulderbreadthcanbeassumedtobe600mm(TransportationResearchBoard,2010),butitmaybenecessarytohaveadoorwayofmorethan1500mminordertoachievetwostreamsatthedoorbecausesomenon-alightingpassengersinthetrainstoodstillnexttothedoor.Adoorwayof1800mmmaymakethispossible.Becausealightingessentiallyrequirestwostreamsofpassengersfromtheseatingspacesonbothsidesofthedoor,havingtwostreamsatthedoorwouldimprove the alighting flow rate.

Secondly,agreaterthesizeoftheverticalgapbetweenthetrainandtheplatformledtoalessflowrate.Alleviatingtheverticaldifferencefrom250mmto50mmwouldimprovearound10%whenboardingisthemain,whereasthealighting-mainrunsshowedaslightimprovementonly.Apossiblereasonforthiswouldbethat,asthetrainfloorisatahigherlevelthanthatoftheplatform,theboardingmovementrequireseachpedestriantolifthis/herbodytoahigherlevelandthisbodymovementcouldslow down the flow rate.

CONCLUSIONAstheapplicationareasofpedestriansimulationmodelsincrease,moreempiricalevidenceisnecessarytosupportthecalibrationandvalidationofmodels.Inthetransportengineeringdiscipline,evidenceonhoweachdesignfactoroftransportinfrastructureaffectspedestrianflowisessential.Theresultsofour experiments provide useful data and insights into the pedestrian movements at the train platform interface.

REFERENCES

Atkins, 2004, Significant Steps – Research, Research commissioned by UK Department for Transport

Harris NG, 2006, Train boarding and alighting rates at higher passenger loads, Journal of advanced transportation, 40(3) 249-263

Transportation Research Board, 2010, Highway Capacity Manual: 2010, Transportation Research Board, Washington

Wiggenraad PBL, 2001, Alighting and boarding times of passengers at Dutch railway stations, TRAIL research School, Delft University of Technology

Including route choice models into pedestrian movement simulation models

Dietmar Bauer, AIT Austrian Institute of Technology, Wien AUSTRIAJonathan Gantner, AIT Austrian Institute of Technology, Wien AUSTRIA

Currentlyroutechoiceinpedestrianmovementsimulationmodels(PMSMs)predominantlyislimitedtotheassumptionofshortestrouteobjective(Kretz2009).Thisneglectsthefactthatpersonsnotfamiliarwiththeinfrastructurehavedifficultiesfindingtheshortestpathandtrytofollowsignage(Schrom-FeiertagandMatyus,2011).Moreoveritisknownthatpedestriansuseavarietyofcriteriaforroutechoice(Golledge1995).

ThispaperprovidesmoreevidencegatheredusingtwoexperimentsconductedduringaworkshopatAITandproposesamodelforroutechoicetoincludethecorrespondingresults.25individuals,allexpertsinpedestrianmodelling,hadtoconducttworoutechoicerelatedtasks.ThefirsttaskconsistedindrawingarouteonamapofthecityofGrazconnectingastartpointwithanendpointwhilepassingeightintermediategoals.Theinstructionsstatedthatallavailableroadscouldbeusedanddidnotinanywayindicatetheobjectivetobeusedforroutechoice.Nosubjectwasfamiliarwiththecity.Outofthe25persons,threepersonsdidnotcompletethetaskcorrectlyleavingasampleof22.

Therouteswerequantizedusinganumberofcriteriatakenfrom(Golledge1995)includinglengthoftheroute,numberofcorners,extentofretracingitsownroute,percentageoflengthonmainroutes(theTeleatlasmapshowsmainroadsinyellow,minorroadsaredrawninwhite),numberoftimes(outofninepossibilitiesfortheeightintermediateandthefinalgoal)thelongestlegoccurredfirstinthecorrespondingroutesegment.Sincetheretraceratioandthenumberoflongestlegfirstdidnotappeartocontaininformationrelevantforclusteringtheindividuals,theywereexcludedfromfurtheranalysis.Afternormalizingeachoftheremainingthreecriteriato[0,1],ak-meansclusteringalgorithmwithfiveclusters(selectedusingthetypicalinformation)wasapplied.Thefiveclustersfoundcanbegiventhefollowinginterpretation:Oneclustercontainingonlytwoindividualsischaracterizedbyexcessiveroutelengthandalargenumberofcorners.Thisclusterwasclassifiedasoutliers.Onecluster(threepeople)ischaracterizedbyahighpercentageofmainroadsandcomparativelyfewcorners.Twoclustersshowtheshortestroutesanddifferinhighusageofmainroads(4people)andsmallnumberofcorners(5people).Theremainingclusterwasthelargest,containingeightpeoplewithnoapparentobjective being followed.

Amoredetailedanalysisoftheirroutesshowedthatthehigherroutelengthwasnotaconsequenceofthechosenorderoftheintermediategoalsbutratherthefailuretofindtheshortestconnectionbetween these.

Thisconjecturewasconfirmedusingthesecondexperimentwhichaskedthepedestrianstowalkalongarouteconnectingfive(outofeleven)givensymbolsrandomlyplacedonthefloorofanotherwise

emptyhall.Herefindingtheshortestpathbetweentwosymbolsistrivial.Nostatisticallysignificantdifferencesinroutelengthsbetweenthelastthreeclusterscouldbefound.

ThisbehaviourmustbereflectedintheroutingcomponentofPMSMs.Theclassicalsocialforcemodel(HelbingandMolnar1995)includesthedefinitionofanotfurtherspecifiedgoalthatneedstobesetbythemodelforeachagentineachtimeframe.Thisleavesroomtoincorporatedifferentagentbehaviourincluding agents unfamiliar with the infrastructure or agents that use full information. In the latter case theresultsabovesuggestthatthepopulationneedstobesplitintotwosegments,agentswhomaximizestayingonmajorroadsandagentswhowanttominimizewalkinglength.

Tothisendagraphoftheinfrastructureisconstructedconnectingintermediategoalsasnodes.Furtheraclassificationintomajoredgesandminoredgesmustbetakenonthebasisofusagecharacteristics(suchasthewidthofcorridor,meansoflevelchanges)aswellassignageinsidetheinfrastructure(edgeswhichcoincidewiththeintendedroutesaccordingtosignagebeingmajorroads).Secondlyalsoapartitioningintomajorandminornodesmustbetaken,wheremajornodesmarkentriestobuildingsorpointsoflevelchangesinsidetheinfrastructure(suchaselevators,escalatorsandstaircases).Allothernodesareseenasminornodes,implementingahierarchyinroutechoicewherefirstchoicesontheupperlevelaremadeandforgivensequenceofmajorintermediategoalsroutechoiceonthelowerlevel is performed.

ThecorrespondingroutechoicecanbemodelledusinganestedC-logitformulationonthisgraphwhichisstandardintrafficflowmodelling.Consequentlygeneraltrafficflowmodellingcanbeappliedincludingroutechoiceundercongestionbyusinguseroptimalassignment.Asforpedestrianinfrastructuresthecorrespondinggraphwillberelativelysmall,implementationofsucharouteisstraightforward.

Thispaperonlyisafirststepintothisdirectioninthesensethatthemainideaisdiscussedandsupportedbyempiricalobservations.Thesamplesizeintheexperiments,however,istoosmallinorder to provide even magnitudes of the percentages of persons using a particular routing strategy. Thusfutureworkwillbedirectedintotheassessmentofthepercentagesofpersonsusingroutelengthminimizationorpercentageonmainroadsastheirobjectivefunctionindifferentrealworldsettings.

ACKNOWLEDGEMENTSTheauthorsthankIrmgardZeiler,ChristianRudloff,StefanSeerandMarkusRayforsupportinthedata collection and analysis.

REFERENCES

Golledge, R. G. 1995. “Defining the criteria used in path selection.” Transportation 278, 151-169Helbing, D., and P. Molnar. 1995. “Social Force Model for Pedestrian Dynamics.” Physical Review E 51: 4282–4286.

Kretz, T. 2009. “Pedestrian traffic: on the quickest path.” Journal of Statistical Mechanics: Theory and Experiment 2009.

Schrom-Feiertag, H., Matyus, T. 2011. „Simulation of handicapped people finding their way through public buildings.” Abstract submitted to the PED2012 conference.

Dynamic medium scale navigation using dynamic floor fields

Dirk Hartmann, Siemens AG, München GERMANYJana Mille, Siemens AG, München GERMANYAlexander Pfaffinger, Siemens AG, München GERMANYChristian Royer, Siemens AG, München GERMANY

Microscopic pedestrian simulations are an emerging topic in all phases of the design and operation ofinfrastructures,e.g.airports.Applicationsrangefromanoptimaldesignofairportswithrespecttopassengercomforttoevacuationofbuildinginthecaseofemergency.Toensureamaximumofreliability of simulations these have to yield as realistic results as possible.

Navigationinexistingmicroscopicpedestriansimulatorscanbedistinguishedintothreedifferentlevels.Onalargescalelevelpedestriansnavigatefromonepointofinteresttothenexttotheirfinaltarget,e.g.fromcrossingtocrossing.Onthislevelotherpedestriansaretakenintoaccountonlyverycoarsely,sincetypicallythewaycannotbeobservedcompletely.Onamediumscalelevel,pedestriansnavigatefromtheircurrentpositiontothenextpointofinterest,whichwewillreferinthefollowingasanintermediatetarget.Dependingonthedensitiesotherpedestriansaretakenintoaccountonthisnavigationlevel,i.e.ifpedestriansseedenscrowdstheyadapttheirrouteimmediatelytryingtomovearoundthecrowd.Onasmallscalelevelpedestriansadapttheirrouttryingnottocollidewithotherpedestrians,whichwewillreferasshortscalenavigation.Here,pedestriansinthedirectneighbourhoodaretakenintoaccount.

Mostpedestriansimulatorstakeotherpedestriansonlyonthelargeandsmallscaleintoaccount.Onthemediumscaleotherpedestriansareneglectedandastaticnavigationfieldisconsidered.Consideringlowtomediumdensitiestypicallygoodandrelativerealisticresultsareobtained.Inthecaseofhighpedestriandensitiesthisishowevernotalwaysthecase.Onlyveryfewsimulatorsusedynamicmedium-scalenavigationtakingotherpedestriansintoaccount(BursteddeKlauckSchadschneider&Zittarz2001,Ketz2009,Hartmann2010).Sincerecentlyeffectivecomputationaltechniqueshavebeenproposedforconstructingappropriatedynamicfloorfields(Hartmann2011),thesimulation of large dens crowd has become well in reach.

Inthispaper,wewilloutlineindetailamodificationofthemethodproposedin(Hartmann2010).Thisallowsthatnotonlyunnaturalcongestionsareavoided,butalsorealisticallycaptureseffectslikelaneformation in dense crowds. Instead of considering a repulsion of pedestrians on a short and medium scaleindependentofthewalkingdirection,themethodstakeswalkingdirectionsintoaccount.Ontheshortscalenavigationissimilarmodelhasbeenrealizedinmanydifferentapproaches,e.g.(Helbing&Molnár1998).Onthemediumnavigationlevel,directionsofotherpedestriansaretypicallynottakenintoaccount.Hereotherpedestriansandtheirwalkingdirectionsaretakenintoaccountbasedonadynamicrecalculationofthenavigationfloorfieldusingpedestrianrepulsionpotentialsadaptedtorelativewalkingdirectionsofpedestrians.

Theapproachdoesnotneedanadditionalfloorfieldasrequiredbytheapproachof(BursteddeKlauckSchadschneider&Zittarz2001).Furthermoretheapproachisasefficientasthemethodsoutlinedin(Kretz2009)and(Hartmann2010),butsimultaneouslycapturestypicaleffectslikelaneformation.Thestrengthoftheapproachisdemonstratedbysimulatingthedifferentscenariosoutlinedin(Helbing,Molnár,Farkas&Bolay201)andsuccessfullyreproducingthedescribedeffectsoflaneformationinmanydifferentsituations.

Usingmoreefficientstrategiesthanthestrategypresentedin(Hartmann2010)formovingalonggradientsofthefloorfiled,aswellasmoreelaboratealgorithmsforcontinuouslyupdatingfloorfields,themethodisnearlyasefficientasclassicalfloorfieldbasedapproachestomicroscopicpedestriansimulators.However,theextensionsofthemodelallowsignificantlymorerealisticsimulationresultsand thus a better prediction of pedestrian flows and egress times.

Thusthemethodisapromisingcandidatetoreplacethemediumscalenavigationlayerinthemulti-levelpedestriansimulatoroutlinedin(KneidlHartmannBorrmann2011)

REFERENCES

Burstedde, C, Klauck, K., Schadschneider, A. and Zittarz, J.. 2001 Simulation of pedestrian dynamics using a two-dimensional cellular automaton. Physica A: Statistical Mechanics and its Applications 295: 507-525

Hartmann, D., 2010. Adaptive pedestrian dynamics based on geodesics. NewJournal of Physics 12 (4), 043032.

Hartmann, D. and Hasel, P. 2011. Efficient floor field methods for microscopic pedestrian crowd simulations. submitted

Helbing, D., Molnár, P., Farkas, I. and Bolay, K. 2001. Self-organizing pedestrian movement. Environment and Planning B: Planning and Design 2001 (28), 361-383.

Helbing, D. and Molnár, P. 1998. Social force model for pedestrian dynamics. Physical Review E 51: 4282-4286

Kneidl, A., Hartmann, D. & Borrmann, A. 2011. Using a multi-scale model for simulating pedestrian behavior, submitted

Kretz, T., 2009. Pedestrian traffic: on the quickest path. Journal of Statistical Mechanics: Theory and Experiment 2009, P03012

Pedestrian gap acceptance in micro-simulation modelling

Paul Townsend, Crowd Dynamics International, Knutsford UNITED KINGDOMLenny Winsel, Quadstone Paramics, Edinburgh UNITED KINGDOM

ThispaperdevelopsanagentgapacceptancebasedalgorithmformodellingopportunisticagentbehaviourandsafercrossingbehaviouratcrossingsusingtheUrbanAnalyticsFramework(UAF).TheUAFsoftwarecombinestheQuadstoneParamicstrafficmicro-simulationmodelandcrowd/pedestrianalgorithmsdevelopedbyCrowdDynamicsInternationalLimited.Uniquealgorithmsfusethesetwoparts,allowingtheindividualvehiclesandagentstointeract.Thisallowsmanydifferentgeometriesandbehaviourstobemodelled,includingpedestriancrossings,sharedspace,pedestrianpresence at signalised junctions and pedestrians who do not comply with signals and cross against aredmanor‚Don‘tWalk‘signal.However,therearealsomanyunmanagedcrossingsinurbanenvironments,wherepedestriansmustmakeadecisionastowhenasufficientgapintrafficexistssothatitissafetocross.Thispaperwillidentifymethodsthatallowthisgapacceptancebehaviourtobemodelled in a flexible manner.

Themethodsdevelopedtomodelpedestriangapacceptanceusetheconceptofscanareas.Eachvehicleisassignedtwopolygonalshapes.Theprimaryscanareawillbetheareawhereavehiclewillbeabletoseepedestriansandtoslowdownforthem.Thesecondaryscanareawillbeanareathatwillallow any agent inside it to consider whether to cross in front of the vehicle depending on combined gapacceptancecriteriadefinedfortheagent,vehicleandlocation.Bothscanareastakeintoaccountindividualvehiclekinematicsandagentattributes.Thiswouldallowagentbehaviouratunmanagedmarkedandrefugecrossingsandmedianstripenvironmentstobemodelledmorerealisticallywithmoreflexibilityondefiningthebehaviours.

Theprimaryscanareainstructsvehiclestoslowdownforanagentwhoiswithinaspecifiedarea.Theprimaryscanareaisanirregularhexagonthatadjoinsthefrontofthevehicle.Itisdefinedusingfourmeasurements:EndWidth(EW),StartWidth(SW),StartLength(SL)andEndLength(EL).

EWisbasedonthetotalnumberoflanesthatthelinkonwhichthevehicletravellinghasonbothsidesoftheroad.EWhasanadditional1mwidthoneithersideoftheroadtoensurethatagentsapproachingacrossingpointwillbeseen.ELisbasedontheadditionalbuffertimesetinthelocationof the crossing.

SWisbasedonthewidthofthevehicle,butcanbeextendeduptotheendwidth(EW)byapercentage.Thisallowsbehaviourtobealteredbyallowingthevehicletowaitforagentsthatareatawideangleofviewfromthedriver,asifthedriverwerelookingleftandright;ortoallowthevehicletocontinuemovinginamoreaggressivestyleofbehaviour;orsomearbitraryparameterbetweenthetwo.

SLisbasedonthevehicle‘sminimumdesiredstoppingdistance.Thisistakenfromthesecondaryscan area determines whether an agent can safely enter the crossing by performing gap acceptance checksagainstconflictingvehicles.Theshapeisatrapeziumthatadjoinsthefrontofthevehicle.Thedimensionsofthisscanareaaredefinedbyfourpointsusingthreemeasurements:Width(W),NearsideLength(NSL)andAdditionalLength(AL).

WisbasedontheEWofthePrimaryScanArea,whichcoversthewholeroadwidthwitha1mbufferto ensure approaching agents are included in the calculations.

NSListhelengthofsecondaryscanareawhenthevehicleisinthelanenexttothekerbline.Thisistheshortestlengthofthedimensionsbecauseitwouldtakeanagentlesstimetotravelpastthevehiclethanit would if the vehicle was in any other lane.

ALisaddedonforeachlanethevehicleisawayfromthekerbline.ThisiscontrolledasapercentageoftheNSLandisaddedonforeachlanethatthevehicleisawayfromthekerbline.TheALisaddedontoboththenearsideofthevehicleandthefarsidedependingonwhichlanethevehicleisin.Agentswillwaitwhenevertheyareinsideasecondaryscanareaandapproachinggapfindingspace.Thisisextremelyusefulfordefiningbehavioursinaspecificareaallowingallagentstotakesimilargaps.

However,theabovemethoddoesnotallowfordifferingbehavioursinagents.Forexampleafastermovingagentmaybemorelikelytotakeasmallergapthanaslowermovingagent.Therefore,anadditionalalgorithmcanbeusedtoallowthemodeltovarythegapchoicefordifferentagents.Thisisdonebycalculatingthetimetakenbyavehicletoreachatheoreticalcollisionpoint.Eachagenthasabufferzonewhichisbasedonthemaximumwalkingspeedofanagent.Thebufferzoneistheadditionaltimewhichneedstobetakenbyanagenttocrossbeforeavehiclearrivesatthecollision point.

Thecollisionpointiscalculatedusingthecurrentlineofsightofanagent.Thisallowsanydirectionofmovementtobetakenintoaccount.Theagentwilllookalongthelineofdirectiontheagentisfacinguntilanavailablewalkingareaisfound.Thisdistanceisusedtoassessifanagentisabletocrosstheroadbeforethevehiclepassesthecollisionpoint.Thisvariableapproachallowsdifferentagentbehaviours to be modelled in the same location.

Inthepaper,theaboveapproachesaredevelopedindetail.Thelimitationsofthealgorithmswithregardstotypeandstyleofcrossingbehaviourareidentifiedandwherepossible,theseareovercomebyallowing a dynamic alteration of the scan area shapes.

A macroscopic multiple species pedestrian flow model based on heuristics implemented with finite volumes

Frank Huth, Technische Universität Berlin, Berlin GERMANYGünter Bärwolff, Technische Universität Berlin, Berlin GERMANYHartmut Schwandt, Technische Universität Berlin, Berlin GERMANY

INTRODUCTIONDespitealltheprincipleimplicationsofapplyingmacroscopicmodelstopedestrianflowsimulations,therearegoodreasonstousethemforsufficientlylargecrows.Theapplicationoffluiddynamicsequationsprovedtobeoflimitedsuccess,sonewapproacheshavebeensought.Onepromisingapproachhasbeenintroducedby(B.PiccoliandA.Tosin,2011)andrelatedandfollowupwork.Anotherapproachhasbeentakenby(S.Berresetal.,2011).Thesepapershaveapartedfocusonmathematical rigor and implementational detail.

Ourmodelhasbeendevelopedwithafocusonpracticalapplicabilityandisintendedtoberatherpragmatic,thanmathematicallyrigorous.Wetrytoovercomeattheonehandthelimitthatonlytwopedestrianspecies(differingindesiredtargetandwalkingspeed)maybehandledasinthetwopapersmentionedaboveandattheotherhandenforceamaximumdensityofthepedestrians(asrequiredinreallifesituations),whichseemsnottobeprovidedbythemodelof(B.PiccoliandA.Tosin,2011).

MODELThemodelshallprimarilysimulatenormal(non-panic)situations.Theprimarygoalsinmindhavebeen:-robustness,-simplicity(asfaraspossible),-standaloneusabilityandintegrabilitywith/intoothermodelsandsoftware,-flexibilityto-provideforadaptabilityintheresearchprocess,-beabletoserveasamoduleintheframeworkoftheMATSimproject(seehttp://www.matsim.org/),-beabletocouplewithamicroscopicmodel,-beingheuristics-basedtoaimatcatchingvitalfeatures,-beingabletomodelapotentiallyunlimitednumberofpedestrianspecies.

Webaseourmodelonthefollowingbasicassumptions:

Assumption1:Pedestrianmovementisdeterminedbyafieldofinfluencesthatresultsintheirwalkingdirectionandspeed.Inthatmodeleveryconsideredaspect(likeplanedpathtoatarget,obstacles,walls,otherpedestriansandsoon)hasafootprintintheenvironmentofthepedestrians,generatinganinteraction

fieldwithdistantforcesorrathereffects,thatcanbeapproximatedcontinuouslyinspace.Theinformationalpeacesofthisfieldcanbeexpressedbypartialdifferentialequations.

Assumption2:Theinformationbase,thatisprocessedbyindividualpedestrianstomakedecisions,isnotpurelyfactual,butaperceptionorevena(re)constructedpicture(basedontheexperiencesoftheseindividuals)ofthereality.Inthecaseofnon-collisiondrivenflow,thevelocityandwalkingdirectionisaproductofaheuristics-baseddecision-makingprocessbyindividualpedestrians.

Assumption3:Theheuristics,thatapplyinpedestriantrafficshouldberathersimpletobeaccessiblebythemajorityofpedestrians.Ifseveralsimplestrategiesaretobeconsidered,theone,thatprovesmoreeffectiveislikelythepreferedone,becausepedestriansaresupposedtoseekefficiencytoo.

Assumption4:Duetothesmoothnessofthecontrollingfields(seeassumption1),weassume,thatamass(inducinginertialbehaviorinthemodel)isnotnecessary.Thiswayweassume,thatthepedestriansmayfollow(adaptspeedanddirection)tothecontrollingfieldswithoutsignificantlagbymeansofinternalimpetus,decisionandphysicalstrength.

Assumption5:Thereexistsafundamentaldiagram,thatexpressestherelationbetweendensity,possiblewalkingspeedand thus flux of the pedestrians.

Thestartingpointofourmodelisderivedfromthemodelin(S.Berresetal.,2011).Themajoradaptation(asidefromdroppingthetwospecieslimit),hasbeentoreplacethe(cross)diffusionterm(whichwedidn‘tfindanappropriateinterpretationfor)byatotaldensitygradienttermandgeneratingatime-dependentdesireddirectionfieldforthepedestrianspecies.

Sothetransportequationsareindividualforeachpedestrianspeciesandarecoupledbythetotaldensity.Thetransportvelocityforeachpedestrianspeciesdissolvesintotwoparts:(1)anintentionaldirectionfolloweddependentonthetotaldensity(accordingtoafundamentaldiagram).ThisintentionaldirectionisthegradientofthesolutiongeneratedbyaPoissonequationanddepends on:-thegeometryofthesimulationarea(globalphenomenon)-ajam-detectionsourceterm(globalphenomenon)-aninter-pedestrian-speciesattraction/repellencysourceterm(localphenomenon)(2)alocalcorrectionvelocitydependentonthegradientofthetotaldensity.

RESULTSTheratherintermediatescalesimulationsproducedtheusuallytobeexpectedqualitativephenomenalikelaneformation,clusteringofpedestriansandcongestionsforappropriatesettingsofparameterswithuptofourinteractingpedestrianspecies.Effects,thataremissingforinstance,areformingofroundaboutsandeffects,thatarebasedonprospectivehumanabilities.

Duetotheconstructionofthemodel,theadherencetotheoverallandindividualpedestrianspeciesdensitieshasbeenensured.Theseresultshavebeenachievedbyafairlysimplesetofrules.

REFERENCES

Piccoli, Benedetto and Tosin, Andrea: „Time-Evolving Measures and Macroscopic Modeling of Pedestrian Flow“, Archive for Rational Mechanics and Analysis, Springer, pp 707-738 (2011)

Stefan Berres and Ricardo Ruiz-Baier and Hartmut Schwandt and Elmer M. Tory: „An adaptive finite-volume method for a model of two-phase pedestrian flow“, Networks and Heterogeneous Media (NHM), AIMS (2011)

A simplified force model and enhanced steering for a quantitative description of pedestrian dynamics

Mohcine Chraibi, Forschungszentrum Jülich GmbH, Jülich GERMANYArmin Seyfried, Forschungszentrum Jülich GmbH, Jülich GERMANYAndreas Schadschneider, Universität zu Köln, Köln GERMANY

Force-basedmodelsareaverypopularapproachformodelingpedestriandynamicswhichassumesthatpedestrian‘smovementisaconsequenceofexteriorforcesactingonpedestrians.Inthisworkweproposeaspacecontinuousforcebasedmodeltodescribethemovementofpedestriansin2D-spacebymeansofforcesandtorques.Furthermore,themodelincorporatesaforesightmechanismallowingpedestrians to react on the actual as well as on a predicted situation.

FormallythemovementofpedestriansisdefinedbyNdifferentialequations.Theequationofmotionisgivenbysuperpositionofrepulsiveanddrivingforces.Whilerepulsiveforceskeeppedestriansawayfromeachotherandotherobstacles,drivingforcesleadthemtoachosenexitwithapreferredvelocity.

Inmostforce-basedmodels[2-11]pedestriansaremodeledascirclesormasspoints.Sincetheforcesactonthecenterofmassofeachpedestrianandgiventhepointsymmetryofpedestrians,thetorquesarezero.Thereforethemovementofpedestriansisrestrictedtoaccelerationsanddecelerationswithouttheabilitytoavoidotherpedestrians.Ingeneraltheseavoidancemaneuversareintroducedinforce-basedmodelinformofalgorithmicsolutions.Inthisworkwemodeltheshapeofpedestriansasellipticaldisksandenhancethegeneralizedcentrifugalforcemodel[1].Withaproperchoiceoftheactingpointoftheforcestakingintoaccounttheactualvolumeexclusion,thedirectionofpedestriansisdetermined,bymeansoftheaforementionedforcesbutalsobytorquesthatthoseforcesproduceintime.

Thus,noextraprocedurestomanagecollisionsoravoidancemaneuversarenecessary.Thisapproachleads to a realistic description of volume exclusion and short termed evasion maneuvers.

For the steering on a tactical level a new method to choose the desired direction is investigated and testedingeometrieswith90°and180°corners.Themethodisbasedonidentifyingautomaticallyinagivengeometrycornersandsetlinesrotatedwithacertainanglearoundthem.Toavoidcongestionsaroundthecorner,eachlineisassignedpointsweighteddecreasinglyawayfromthecorner.Incaseoflowdensities,pedestriansareguidedtowardsthecorner.Ifthedensitygetshigh,pedestriansget,thanksoftheweights,directedawayfromthecorner.

Qualitativeandquantitativecomparisonsamongseveralexperimentaldataandmeasurementmethodsare used to validate the steering model.

Themodelforthevolumeexclusionandsteeringcontainsonlytwofreeparameters,whichbenefitsthisprocedure and facilitates its calibration.

Forthesakeofvalidationofthemodelforthevolumeexclusionwereproduceempiricaldatainseveralgeometrieswithonesetofparameter.TheGCFMwassuccessfullyvalidatedinnarrowandwidecorridors.Forbottlenecks,corners,T-junctionsandgenerallygeometriesthatnecessitatemaneuvering,theneedtomodelthedesireddirectionisespeciallymorehighlighted.Withtheenhancements,weintroduceinthiswork,itispossibletosimulatemorechallenginggeometrieswithouttheneedoftuning and changing parameters.

In summary the proposed model reflects three aspects:

1.Therepulsiveforcereflectstheellipticalshapeofpedestrians.Itsmagnitudeisproportionaltotheoverlapping area with other pedestrians.

2.Therepulsiveforceengenderatorque,allowingpedestrianstochangeeasilydirection.

3.Theforcesdependnotonlyontheactualsituationbutalsoonaneducatedguesshowthesituationchanges in the future.

4.Thedesireddirectionismodeledtoenablemaneuveringincomplexgeometries.

REFERENCES

[1] M. Chraibi, A. Seyfried, and A. Sachdschneider. Generalized centrifugal force model for pedestrian dynamics. Physical Review E, 82:046111, 2010.

[2] Z. Fang, J. P. Yuan, Y. C. Wang, and S. M. Lo. Survey of pedestrian movement and development of a crowd dynamics model. Fire Safety Journal, 43(6):459 465, aug 2008.

[3] D. Helbing and P. Molnar. Social force model for pedestrian dynamics. Phys. Rev. E, 51:4282 4286, 1995. [4] T. I. Lakoba, D. J. Kaup, and N. M. Finkelstein. Modifications of the Helbing-Molnr-Farkas-Vicsek Social Force Model for Pedestrian Evolution. Simulation, 81:339 352, 2005.

[5] Rainald Löhner. On the modelling of pedestrian motion. Applied Mathematical Modelling, 34(2):366-382, feb. 2010. Article in Press.

[6] Daniel R. Parisi, Marcelano Gilman, and Herman Moldovan. A modification of the social force model can reproduce experimental data of pedestrian flows in normal conditions. Physica A: Statistical Mechanics and its Application, 388(17):3600 3608, sep 2009.

[7] A. Seyfried, B. Steffen, and T. Lippert. Basics of modelling the pedestrian flow. Physica A, 368:232 238, 2006.

[8] B. Steffen and A. Seyfried. The repulsive force in continous space models of pedestrian movement. 2008. arXiv:0803.1319v1.

[9] B. Steffen and A. Seyfried. Modelling of Pedestrian Movement around 90 and 180 Bends. In B. H. V. Topping and Y. Tsompanakis, editors, Proceedings of the First International Conference on Soft Computing Technology in Civil, Structural and Environmental Engineering. Civil-Comp Press, Stirlingshire, UK, 2009,, 2009.

[10] Albert Steiner, Michel Philipp, and Alex Schmid. Parameter Estimation for a Pedestrian Simulation Model. In Swiss Transport Research Conference, 2007.

[11] W. J. Yu, L.Y. Chen, R. Dong, and S.Q. Dai. Centrifugal force model for pedestrian dynamics. Phys. Rev. E, 72(2):026112, aug. 2005.

Quantitative and qualitative validation for general use of pedestrian models

Mario Campanella, Delft University of Technology, Delft NETHERLANDSSerge Hoogendoorn, Delft University of Technology, Delft NETHERLANDSWinnie Daamen, Delft University of Technology, Delft NETHERLANDS

INTRODUCTIONTypicallytheapplicationofamodelincludesadescriptionofthepopulationcharacteristics,thewalkinginfrastructureandinflows.Thepopulationisusuallyrepresentedinternallyinthemodelanddetermined via the model parameters that were obtained in an optimisation procedure during the calibrationprocess.Theinfrastructureandinflowsdeterminethetypeofsituationspedestrianwalk.

Thefundamentalpremisefortheuseofpedestrianmodelsistheconfidencethattheirresultsarewithinanaccepteddeviationofreality.However,pedestrianbehavioursvaryaccordingtolocalconditionssuchasdensities,directionofinteractionbetweenpedestrians(walkingtowardseachother,crossing,overtaking)andinternalstatesdeterminedbyage,culture,purposeofthetrip.Theprocessofdeterminingtheadequacyofamodel(anditsparameterset)iscalledvalidation.Atypicalvalidationwill compare some empirically obtained data about pedestrians with the outcomes of the model. Ideally thecomparison(assessment)isperformedreproducingsimilarpopulationcharacteristics,walkinginfrastructureandinflows.Aliteratureresearchonvalidationofpedestrianmodelsshowedthatusuallyauthors perform a limited amount of validation assessments and very few considerations are given regardingtheapplicabilityofthemodelindifferentwalkingsituationsandpopulationcharacteristics.Wearguethatifnoparametersetiscalibratedforthespecifictask,thesettobeusedmusthaveshowntoperformwellinvariousvalidationassessmentsutilisingseveralwalkingsituationsgivingameasureof general usability.

Thispaperproposesasimpleandmeaningfulcriteriontocombinequalitativeandquantitativevalidationassessmentstoobtainameasureofvalidationquality.ApplyingthismethodwiththemicroscopicNomadmodeldevelopedbytheTechnicalUniversityofDelft,resulted(asexpected)thatparametersetsthatwerecalibratedwithflowsthatexposepedestrianstodifferentsituationsandoptimisedwithseveralcalibrationassessments(multi-objectives)aremoreaccurateandpresentmorerealistic pedestrian behaviour.

METHODOLOGYThevalidationendswithageneralscoreobtainedaftercombiningtheresultsofalltheassessments.Eachassessmentresultedinascorethatrangedbetween,bad,mediumandgood.Thequantitativeassessmentsweregradedaccordingtothesizeoftherelativeerrorsofthesimulatedresults:bad:=error>10%medium:=5%<error<10%good:=error<5%.Eachqualitativeassessmentwasassignedoneofthescoresaccordingtotheirresemblancetotheempiricaldata.Thecombinationofthescoresisdoneusingasimpleaveragingprocedureperformedaftergradingthescores.Thebadscoregetsagrade0,themediumgets1andthegoodgets2.Anaverageofthegradeswasthentabulatedbetween

threeintervalsofequallengthtocalculatethefinalscore.Thisverysimpleprocedurewaschosentodiminish the complexity of the validation procedure allowing for a simple interpretation of the results. Toincreasetheimportanceofthequantitativeassessmentsinthefinalscoreallqualitativeassessmentswerecombinedintoonescoreandthenaddedtoallquantitativeassessmentstoobtainthefinalaveragegrading.

Thevalidationassessmentsperformedinthispaperuseddatafromthreetrajectorysetsobtainedincontrolledwalkingexperiments(Daamen2003).Thesesetscoverdifferenttypesofflow:abidirectionalcorridor,aunidirectionalcorridorandaunidirectionalflowwithnarrowbottleneckcorridor.Theseflowsrepresentmanywalkingsituationsandaremostoftenusedinvalidationassessments.

Thethreemostcommontypesofvalidationassessmentsfoundintheliterature,traveltimes,speedxdensityfundamentalrelationandbottleneckcapacityestimationwerequantified.Thequalitativevalidationjudgedtheinteractionandwalkingbehaviours.Furthermore,forthebidirectionalflowwequalifiedthedurationandstabilityoftheself-organisedlanes.Thenarrowbottleneckexperimentspresentedtwodistinctivesituations,upstreamofthebottleneckcongestionbuildsupformingafunnelofpedestrianswiththeapexlocatedinthebottleneckanddownstreamwherepedestrianswalkinsideanarrowcorridorinastaggeredpositioningalsoreferredastheselforganizingzippereffect.Bothphenomenawerealsousedforthequalitativevalidation.

Wecomparetheresultsofeightparametersetsthatwerecalibratedwithdifferentobjectivesovertrajectorysetsrepresentingthethreeflows.Twoobjectivefunctions,thetraveltime’serrors(TT)andtheaccelerationerroratwalkingsteps(Ac)wereusedwiththethreetrajectorysetstotallingthesixspecialisedparametersets.Theothertwosetsaregeneral;themultiTTwasoptimisedusingamulti-objectivefunctionthatcombinedthetraveltime’serrorsofthethreetypesofflow.Themultiisevenmoregeneralbecauseitwasoptimisedwithdifferentobjectivefunctionsforthethreeflows.Thecharacteristicsofeachflowwereexploredinthecalibration,thequantityoflanes,thespeedxdensityrelationandthebottleneckcapacity.Eachassessmentwasperformed30timestoaccountforthestochastic variations of the model and the results were averaged.

RESULTSANDDISCUSSIONThefirstimportantresultwasthatthestochasticvariationoftheassessmentsusingthegeneralparametersetswassignificantlysmaller.Thereasonforthiswasfoundinthesignificantlybetterbehaviourofthepedestrians.Bothgeneralsetsobtainedbetterqualitativescoresthatmeasuredtheinteractions,themultisetobtainingthemaximumscoreinallthreeflows.Theoverallscoreofthegeneralsetswereidenticalwithonegoodandtwomediums,alwaysbetter(orequal)thenthespecialisedsetsforalflows.Consideringthatthespecialisedsetswerevalidatedtoflowsidenticaltothoseusedintheircalibration,theseresultsareveryimportant.Theyindicatethatpedestrianbehavioursaresocomplexthatthecalibrationusingonetypeofflowandoneassessmentisoverfittingtheparametersandthatcombiningseveralaspectsofwalkingbehavioursapproximatebettertoaglobaloptimumofpredictions.Theunidirectionalflowobtainedthebestvalidationresults(allsetsobtainedamedium)andalsothebestqualitativereproductionofindividualbehaviours(twogoodsandtwomedium).Thespeedxdensityassessmentgavetheworseresultsforallsetsandallflows.Thiswasmostlyduetothedifficultyofachievingthesamemaximumdensitiesinthesimulationsofthebidirectionalandtheunidirectionalflowsandthelargedatascatter.Thesevariationsinthevalueoftheassessmentsclearlyshowtheindependenceofthem.Presentinggoodresultsforoneisnoguaranteethat the same will happen with the other assessments.

CONCLUSIONThispaperproposedanewandverysimplevalidationprocedurethatcombinesqualitativeandquantitativeassessments.Spiteitssimplicityitwasabletoshowthatseveralassessmentseffectivelyrevealthedifferencesinaccuracybetweenparametersets.Thegoodresultsforthevalidationoftheunidirectional flow when the same did not happen for the others show that this type of flow should not be relied solely in a validation. It should always be accompanied by other type of flows that are more challengingforpedestrianmodels.Setsoptimizedwithmultiple-objectivessimulatebetterwalkingbehavioursanddiminishthevariabilityoftheresults.Thesmallervariabilityandthebetteraccuracymakethemmoresuitedtobeusedinsimulationsandshouldaccompanyallpedestrianmodels.Theseresults strongly encourage the pursuing of accuracy in a wide range of situations (generality).

REFERENCE

Daamen, W. and Hoogendoorn, S. (2003). Controlled experiments to derive walking behaviour. European Journal of Transport and Infrastructure Research, 3(1):39–59.

FDS+Evac model validation for seated row arrangements

Naveesh Reddy, PM Dimensions Pvt. Ltd., Gandhinagar INDIAAshok Babbar, PM Dimensions Pvt. Ltd., Gandhinagar INDIATimo Korhonen, VTT Technical Research Center of Finland, Esbo FINLAND

TheopensourcesoftwareFireDynamicsSimulator(FDS,version-5.5.3)isoneofthemostadvancedcomputationalfluiddynamicsbasedfiresimulationsoftwareavailabletomodelpracticalfireproblems.FDShasgotanotherpowerfultoolwhichiscapableofsimulatinghumanegressprocesswithandwithouttheeffectsoffireunderthenameofFDS+Evac.ThehumanmovementalgorithmofFDS+Evachasbeenvalidatedwithexperimentalevacuationdataandotherevacuationmodels.Butitisnotyetvalidatedonhowtomodelevacuationprocessforseatedrowarrangementslikeaircraft,auditoriumetc.Hereanattempthasbeenmadetomodelevacuationofhumansinseatedrowarrangementswiththehelpoftwoknownexperimentaldata,i.e.,evacuationstudiesofablendedwingbodyaircraft(BWB)andacinematheatre.ThefiredrillevacuationtestsweremodelledusingFDS+Evacandvalidated with the experimental data and other human egress models.

Ablendedwingbodyaircraft(BWB)configurationconsistsof1020passengers,with25cabincrewsand20floorlevelType-Aexits.TheexitspresentontherighthandsideoftheaircraftfromfronttoreararedenotedasR1toR10andonthelefthandsideasL1toL10respectively.Thetestcaseconsideredherewasastandardevacuationcertificationcasewherehalfoftheusableexitsareblocked,i.e.,10outof20exitsweremadeavailableontheleftside(L1toL10)oftheaircraftandastandardopeningtimeof11.1secforType-Aexitsisused.Consideringfull-scaletrailsinvolvingover1000peopleisexpensive,soGaleaetal.[E.R.Galea,L.Filippidis,Z.Wang,P.J.Lawrence,andJ.Ewer,Evacuationanalysisof1000+seatBlendedWingBodyaircraftconfigurations:ComputerSimulationsandFull-ScaleEvacuationExperiment,PedestrianandEvacuationDynamics,Springer,pp.151-161,2011.]decidedtoundertakefullscaletrailsusingaportionofBWBcabin.Atotalof375passengersevacuatedwithaccesstoexitsL6toL10ontherearsideoftheaircraft.Some88participantswereseatedinthemockupand146participantswerebebroughtintothemock-upsectionviatwocrossaislesfeedingthemock-upsection.

SincetheBWBcabinlayoutinformationwithinternaldimensions,aisleswidth,seatdimensionsetc.arenotavailableinthatpaper,thecabinwasmodeledonlyusingtheavailableinformationoftheaircraftmodelpresentedintheirpaper.Initially,theBWBlayoutconsideredinthisworkhasthespaceinfrontofdoorsas1.0m,verticalaisleswidthas1.0mandhorizontalaisleswidthas1.0m.Atotalof11BWBtestcases(BWB,BWBTC1toBWBTC10)areconsideredbyvaryingtheabovedimensionsfrom1.0mto1.25m,and1.5mandevacuationtimeiscalculatedforallthecases.Itwasnoticedthatthevariationofhorizontalaisleswidthdidnoteffectedtheevacuationtimemuch.ThefinallayoutacceptedisBWBTC6,i.e.,spaceinfrontofdoorsas1.5m,verticalaisleswidthas1.5mandhorizontalaisleswidthas1.0m.ThepredictedaverageexitusagefromFDS+EvaciscomparedwithAir-Exodusmodelandithasshownsimilartrendsinexitusages.ThemodeledBWBTC6layoutiscomparedto

theexperimentandthetrendsofpredictedaverageexitusagelooksimilar.Wehavenoticedthattheaverageexitusageofpassengersismorefortheexit-L10inFDS+Evacandmoreforexit-L6inAir-Exodusmodelwhencomparedtoexperimentaldata.SincethepassengersaremovingfromtherightwingtoleftwingtheychoosethenearestexitwhichistheexitL10.Theexit-L7,whichispresentatthebackcorner,istheleastusedbythepassengersandbothmodelspredictedthiswell. Thecinematheatreconsistsof135participantswithtwoexitsoneatthefrontsideandoneatthebackside.Itconsistsofnicerowswithfifteenseatseach.Thespaceforhumanmovementispresentontheeachsideoftheseatrowsasstepsfromfronttoback.Aseriesofevacuationexperimentswereconductedinacinematheatretoinvestigatethesocialinfluenceduringemergencyevacuation.Thepurposeoftheseexperimentswastotesthowdifferentalarmseffectthepre-movementtime.ThedifferentalarmstestedbyNilssonandJohansson[D.NilssonandA.Johansson,Socialinfluenceduringtheinitialphaseofafireevacuation-Analysisofevacuationexperimentsinacinematheatre,FireSafetyJournal44,pp.71-79,2009.]werealarmbellandpre-recordedhumanvoice.Thepre-movementtimeswerealsorecordedintheexperiments.TheaboveexperimentalcasewasalsovalidatedusingSTEPSandSIMULEXmodels.

Thelayoutinformationofthecinematheatreis11.5mx10mwithtwo0.8mwideexitsand1.0mfreespaceforhumanmovementoneithersideoftheseatrows.Thespaceinbetweentheseatrowsis0.5manddimensionsofseatarealsoknown.ThecinematheatremodelwasdevelopedfromtheaboveknowninformationusingFDS+Evac.Theexperimentalpre-movementtimeisgivenasinputtotheFDS+Evac.WehavenoticedthattheFDS+Evacisabletopredictthetotalevacuationtimenicelyalongwithothermodels.TheFDS+Evacmodelpredictiononnumberofhumanspresentinrows2to9atthetimes40,50and60secafteractivationofalarmareclosetoexperimentaldata.

TheFDS+Evacmodelisvalidatedforseatedrowarrangementsbyconsideringtwoexperimentalevacuationcases.FDS+Evacisabletopredictwellforboththecases.Thoughmuchinformationonthepre-movementtimeswasnotprovidedinthecaseofaircraft,itwasnoticedthatthemotiveforceforhumanevacuationinaircraftisfasterwhencomparedtothecinematheatre.TheFDS+Evac,PERSnamelist,whichtakescareofinputparameterslikehumanspeeds,pre-evacuationtimesandsocialinfluence forces etc. for the aircraft and the cinema theatre cases are presented in this paper.

Keynote


Uwe Hanebeck, Karlsruher Institut für Technologie KIT, Karlsruhe GERMANY Multimodal extended range telepresence in pedestrian simulation

Extendedrangetelepresenceprovidesanintuitivewaytoexplorerealremoteorvirtualenvironments.Thefeelingofpresenceisachievedbyvisual,acoustic,andhapticsensoryinformation recorded from the target environment and presented to the user on an immersive display.Inordertousethesenseofmotionaswell,whichisespeciallyimportantforhumannavigationandpathfinding,theuser’smotionistrackedandtransferredtotheteleoperator,e.g.,amobilerobotoravirtualavatar,inthetargetenvironment.Asaresult,inextendedrangetelepresencetheoperatorcanadditionallyuseitsproprioception,i.e.,thesenseofmotion,tonavigatetheteleoperatorbynaturalwalking,insteadofusingdeviceslikejoysticks,pedals,orsteering wheels.

Withoutfurtherprocessingofthemotioninformation,themotionoftheoperatorisrestrictedtothesizeoftheuserenvironment,whichislimited,forexample,bytherangeofthetrackingsystemortheavailablespace.MotionCompressionsolvesthisproblembymappingthedesiredpathinthetargetenvironmenttoafeasiblepathintheuserenvironmentbyminimizingproprioceptive and visual inconsistencies.

Byconnectingthisextendedrangetelepresencesystemtoapedestrianandvehiclesimulationsoftware,dataaboutpedestriandynamicscaneasilybecollectedinexperimentsinwhichnotallparticipantsneedtoberealpeoplebutinsteadsomeofthemcanbesimulated.Thisnotonlyallowsthecollectionofdataaboutpedestriandynamicsbutalsotocalibratemodel-specificparameters.

Inthistalk,anoverviewofthebenefitsofextendedrangetelepresence,forexample,forinvestigatingroutechoicebehaviorinevacuationsituationsandforfindingrealisticparametersinpedestriansimulationsisgiven.Moreover,thistalkpresentsrecentadvancesinextendedrange telepresence including haptic guidance methods and a wireless user environment.

Oral presentations


A.10 Matthias Plaue, Technische Universität Berlin, Berlin GERMANY On measuring pedestrian density and flow fields in dense as well as sparse crowds A.11 Stéphane Bonneaud, Brown University, Providence RI USA An empirically-grounded emergent approach to modeling pedestrian behavior

A.12 Sebastian Burghardt, Bergische Universität Wuppertal,Wuppertal GERMANY Fundamental diagram of stairs: Critical review and topographical measurements of density and flow

B.10 Kevin Rio, Brown University, Providence RI USA A data-driven model of pedestrian following and emergent crowd behavior

B.11 Christian Rudloff, AIT Austrian Institute of Technology, Wien AUSTRIA Comparison of different calibration techniques on simulated data

B.12 Michael Schultz, Technische Universität Dresden, Dresden GERMANY Individual assessment of conflict prediction and group dynamic behavior as main driver for pedestrian dynamics

On measuring pedestrian density and flow fields in dense as well as sparse crowds

Matthias Plaue, Technische Universität Berlin, Berlin GERMANYGünter Bärwolff, Technische Universität Berlin, Berlin GERMANYHartmut Schwandt, Technische Universität Berlin, Berlin GERMANY

Intheframeworkofmacroscopichumancrowdmodels,pedestriandynamicsaredescribedvialocaldensityandflowfields.Intheoryatleast,thesedensityandflowfieldsareoftenrequiredtohaveacertaindegreeofregularitysuchasbeingsmooth.Empiricaldataofhumancrowdbehaviors,ontheotherhand,areusuallyrepresentedbythepedestrians‘trajectories.Probablythemostbasicwaytocompute a density from such trajectories would be to divide the number of pedestrians in a given regionbytheareaofthatregion,atagivenpointintime.However,this„standard“densityestimatoryieldsdatawithlargescatter-letaloneasmoothdensityfunctiondefinedateverypoint.Verysimilarproblems occur when estimating the flow by counting pedestrians passing through a given cross section.

Atleasttwoapproachesformeasuringthe(local)densityhavebeensuggestedintheliteratureasalternatives:-In(Helbingetal.,2007),alocaldensityfieldiscomputedviathesumofGaussianswithfixedstandarddeviation(typically0.7m)centeredateachpedestrian.Thisapproachmayberecognizedasakerneldensityestimationwithfixedbandwidth,whichisabasictoolinstatisticaldataanalysis(see(Silverman,1986),forexample).Thismethodresultsinasmoothdensityfielddefinedateverypoint.Ofcourse,thekernelestimatoryieldsthesameresultasthestandarddensitywhenspatiallyaveragedacrosslargeregions.However,foranareaof„mesoscopic“sizeonetypicallyobservesvaluesthataresignificantlylowerthanthestandarddensitysincealargeportionofthe„pedestrianmass“islocatedoutside of the respective region.-(SteffenandSeyfried,2010)proposetwosimilarestimatorsbasedontheVoronoidiagramdefinedbyeachpedestrian‘spositionasaVoronoisite.Themainideainthisapproachistoaccountforthepersonalspaceoccupiedbyeachpedestrian,andthispersonalspaceisrepresentedbytheareaofthecorrespondingVoronoicell.ThevaluesfortheVoronoidensityareveryclosetostandarddensities,butwithsignificantlylessscatter.However,theVoronoiestimatordoesnotyieldasmoothlocaldensitydefinedateverypoint.Also,forverysparseandunconstrainedcrowds,alargenumberofVoronoicellsmayfailtohavefinitearea,leadingtoconceptualissues.

Inthispaper,wedescribeamethodforcalculatingalocaldensitybasedonkerneldensityestimationwithvariablebandwidth.Themethodissimilartotheestimationtechniquewepreviouslydescribedin(Plaueetal.,2011),butyieldsdensitieswhicharealsotemporallysmoothandaccountformultiplenearby pedestrians as an influence on personal space.

OuralgorithmisconceptuallyablendoftheVoronoiestimator(accountingforpersonalspace)andthefixed-bandwidthkernelestimator(yieldingsmoothdensityfields).Itmaybesummarizedasfollows:

(1)Foreachpedestrian,calculateakindofweightedaverageofthedistancestoeachoftheotherpedestrians.Thelargerthenumberofnearbypedestrians,thelowerthisvaluewillbe;thismaybeseenasamodelforthecompressionandrelaxationofeachindividual‘spersonalspace.(2)Foreachpedestrian,usethevaluecomputedinStep(1)asthebandwidthofaGaussiancenteredatthepedestrian‘sposition.TheGaussiansarenarrowerinacrowdedneighborhood,andthereforelesspedestrianmassislocatedoutsideofdensemesoscopicregions.Thisinturnleadstoabetteragreementwith the standard density estimator for such regions.(3)ObstaclesandwallsmaybemodeledbymultiplyingtheGaussianwithamollifiedcharacteristicfunction with respect to the total observational area.(4)Renormalizeandsumthe(modified)Gaussianstoobtainthelocaldensity.

Furthermore,asanotherextensionofourpreviouswork,wealsospecifyhowtocalculateacorrespondingflowfield:(5)Computetheweightedsumofthepedestrians‘velocitiesviathe(modified)Gaussianscomputedabove.Thisyieldsapedestrianmassflowcomponentsimilartothelocalvelocityestimatordescribedin(Helbingetal.,2007)butwithvariablebandwidth.(6)Computeanadditionalirrotationalflowcomponentsuchthatthecontinuityequationissatisfied.ThistaskisachievedbynumericallysolvingaPoissonequation.Ifobstaclesorwallsarenotaccountedfor,thisflowcomponentmaybecomputedexplicitlyinaclosedform.Fromthisclosedformitmaybereadily seen that the additional flow component is best interpreted as the transport of pedestrian mass via the compression or relaxation of personal space.(7)Thesumoftheflowcomponentscomputedin(4)and(5)representsthetotalflow.

Inordertoevaluateourapproachandcomparethedifferenttechniques,weuseadatasetofintersecting pedestrian flows extracted from human crowd experiments that we conducted at the TechnischeUniversitätBerlin.Inoneparticularexperiment,twopedestrianflows(142and83subjects)intersectatanangleof90degreesforoneminuteinaregionofabout25squaremeters,reachingpeakdensitiesofaboutfourpedestrianspersquaremeter.

Wedemonstratethattheestimationtechniqueproposedbyushasthefollowingadvantages:-Thedensityandflowfieldarerepresentedbysmoothfunctionsdefinedateveryspatio-temporal point.-Thedensityandflowfieldsatisfythecontinuityequation.-Thedensityestimatortypicallyyieldsvaluesclosetothestandarddensity.

Finally,wearguethattheproposedmodelmaybeinterpretedastonotonlydescribethetransportofpedestrianmassviaparticleflowbutalsoastheresultofvariationsinthepedestrians‘personalspaceincrowdedsituations.Wespeculatethatthisapproachmayleadtonewwaysindescribingcrowddisasters which may be thought of as events where a sudden compression of personal space occurs.

REFERENCES

D. Helbing, A. Johansson, and H. Z. Al-Abideen: „Dynamics of crowd disasters: An empirical study“, Phys. Rev. E 75, pp. 046109 (2007).

M. Plaue, M. Chen, G. Bärwolff, and H. Schwandt: „Trajectory extraction and density analysis of intersecting pedestrian flows from video recordings“, Proc. PIA11, LNCS 6952, 285-296 (2011).

B. W. Silverman, „Density Estimation for Statistics and Data Analysis“, Chapman and Hall (1986).

B. Steffen, A. Seyfried: „Methods for measuring pedestrian density, flow, speed and direction with minimal scatter“, Physica A 389, pp. 1902-1910 (2010).Frantzich, H., En modell för dimensionering av förbindelser för utrymning utifrån funktionsbaserede krav, Tech. rep., Department of Fire Safety Engineering, Lund University, (In Swedish) (1994).

An empirically-grounded emergent approach to modeling pedestrian behavior

Stéphane Bonneaud, Brown University, Providence RI USAKevin Rio, Brown University, Providence RI USAWilliam H. Warren, Brown University, Providence RI USA

Acriticalissueforpedestrianandcrowdsimulationisthelackofempiricalvalidationofthebehaviorallocomotionmodel.Theabilitytoaccuratelyreproducepedestrianlocomotordynamicsisnecessaryforcrowdmodeling,urbanplanningordisastermanagement.But,relyingonsimulationrequiresrealisticmodelsofpedestrianlocomotorbehavior.Mostexistingpedestrianmodelshavebeenbasedonad-hocrulesofinteractionandparameters,orontheoreticalframeworkslikephysics-inspiredapproachesthatarecognitivelyquestionable.Furthermore,itisextremelydifficulttovalidateamodelinallpossiblescenarios.Hence,wearguewithothersthatamorecognitivescienceapproachisneeded(Moussaidetal.2011).Andweproposehereanempirically-groundedmodelofhumanlocomotionmotivatedbythecognitively-plausibleecologicalapproachtoperceptionandaction(Gibson1986).Tobuildareliablebehavioralmodel,onemayeitherexperimentallyderivethelocalcontrollawsforlocomotorbehavior,yieldingagenerativemodel,ormeasureandtrytoreproduceglobalcrowdparameters.Yet,thoughamodelmightreproducecrowdparameters,individualstrajectoriescanstillbeunrealistic.Wearguethatagenerativemodelwillnotonlyproduceaccuratetrajectoriesattheindividuallevel,butisalsolikelytoyieldrealisticemergentcrowdbehavior.

Inthispresentation,wefirstlaydownthetheoreticalframeworkoftheapproach.Then,wedescribethecontrollawsoffourelementarylocomotorbehaviors.Finally,weshowhowthemodelcanbeusedinagent-basedsimulationsofcomplexscenarios.Westudyhowourmodelbehaveswithmanystationaryobstaclesandmanyinteractingagents.Wediscusstheperformancesofourmodel,intermsofbehavioralpatternsandcomputationalperformances,inthecontextofcomputersimulationandanimation.

Basedonthebehavioraldynamicsframework(Warren2006),webuiltalocomotionmodelthataccounts for how a human steers towards a stationary or moving goal and avoids stationary or moving obstacles(Warren2008).Theapproachisemergent,i.e.thelocomotortrajectoryisnotdescribedbyaninternalplanningprocess,butemergesfromtheinteractionsoftheindividualagentwithitsenvironment.Ourtheoreticalframeworkistheecologicalapproachtoperceptionandaction(Gibson1979),wheretheindividualiscoupledtoitsenvironmentthroughcontrolvariablesandbehavioralstrategiestocontrolitsactions.Basedonthedynamicalsystemsapproachtoaction(Kugler&Turvey,1987;Kelso1995),interactionsaredescribedwithdynamicalsystemsandthelocomotiondynamicsisthereforeacontinuouspathinthestatespaceofthesystem.Toinferandvalidatethehypothesesontheinformationandbehavioralstrategiesusedbyhumanstolocomoteinspace,weuseobservationsdoneinrealofvirtualworldsthroughcontroledexperimentsinvolvinghumanparticipants.Wethenyieldcontrollawsoflocomotion,whichspecifybehaviorintermsofadynamicalcontrolofperceptualvariables.Boththetheoryandtheempiricalobservationsenabledustodescribeaparsimoniousmodel

using the environmental information and behavioral rules that humans seem to use when locomoting in an environment.

Weproposetodecomposelocomotionintoasetofelementarybehaviorsthatcanbemodeledindividually.Asafirstapproximation,theseinclude(a)steeringtoastationarygoal,(b)avoidingastationaryobstacle,(c)interceptingamovingtarget,and(d)avoidingamovingobstacle.Ourstrategy(Fajen&Warren2003)istomodeleachelementarybehaviorasanonlineardynamicalsystemandthen attempt to predict human behavior in more complex environments by linearly combining these components.Basedonhumanlocomotionexperiments,wedeterminedthebehavioralvariablesastheheading(φ)ordirectionoftraveloftheagent(allocentricreference),andthecurrentturningrate(φψ),assumingforthemomentaconstantspeedoftravelv.Thesimplestdescriptionofsteeringtowardastationarygoalisfortheagenttobringthetarget-headingangletozero(β=φ−ψg=0)asitmovesforward,whichdefinesanattractorinstatespaceat[φ,φψ]=[ψg,0].Conversely,forastationaryobstaclethatliesinabearingdirection(ψo)withrespecttothereferenceaxis,atadistancedo,thesimplestdescriptionofobstacleavoidanceistomagnifytheobstacle-headingangle(φ−ψo>0),whichdefinesarepellerat[φ,φψ]=[ψo,0].Steeringtowardamovinggoalisageneralizationofthestationarycase,andthegoalfortheagentistokeepaconstantbearingwiththetargetasitmovesforward.Conversely,foramovingobstacle,theagentfollowstheinversestrategybyavoidingaconstantbearing with the obstacle.

Aftershowingthedynamicsofthemodelinelementaryscenariosofinteraction,weshowthattheexperimentally-groundedcontrollawsgeneralizetomorecomplexscenarios.Westudyfiveclassicscenarios.Threefocusoninteractionsbetweenpedestrianswalkinginacorridor:(1)Twopedestrianswalkingtowardseachother,(2)onepedestrianwalkingtowardsagroup,and(3)twoflowsofpedestrianswalkinginoppositedirections.Thelasttwoscenariosfocusonapedestrianfindingitswayonacrowdedplazawith(1)Nstationaryobstaclesand(2)Magentswalkinginvariousdirections.Foreachscenario,weshowhowourmodelgeneratesrealisticindividualdynamicsinrespecttotheresultsobtainedintheelementaryscenarios.Scenario3alsoshowshowourmodeliscapableofproducingself-organizationphenomenalikelaneformation.Andthetwolastscenariosshowhowcrowddynamics can emerge based on our simple control laws.

Thismodel,basedonempiricaldataandmotivatedbyacognitivelyplausibletheory,generatesindividual trajectories that accurately match elementary human behaviors. It is not only reliable at theindividuallevel,butalsoproducesrealisticemergentcrowdbehavior.Theaimistoempiricallydeterminetheminimumsetofcomponentsneeded,butadditionalcomponentsmayberequiredtoaccountforbehaviorsuchasherding.Wediscussourcurrentinvestigationsofnewcontrollawsforunderstandingcollectivebehaviorthroughnewexperimentsonleader-followerinteractionsandsmallgroupsofpedestrianswalkingtowardsacommongoal(Bonneaudetal.2011).

REFERENCES

Bonneaud, S., Rio, K. & Warren, W. H. (2011). Accounting for patterns of collective behavior in crowd locomotor dynamics for realistic simulations. Transactions on Edutainment (IN PRESS).

Fajen, B. R., & Warren, W. H. (2003). Behavioral dynamics of steering, obstacle avoidance, and route selection. Journal of Experimental Psychology: Human Perception and Performance, 29(2), 343-362.

Gibson, J. (1986). The Ecological Approach to Visual Perception. New York, NY: Psychol- ogy Press.

Helbing D., Farkas I. & Vicsek, T. (2000). Simulating dynamical features of escape panic. Nature, 407, 487-490.

Kelso, S. (1995). Dynamic Patterns: The Self-Organization of Brain and Behavior (Complex Adaptive Systems). Cambridge, MA: The MIT Press.

Kugler, P.N., & Turvey, M.T. (1987). Information, natural law, and the self-assembly of rhythmic movements. Hillsdale, NJ: Lawrence Erlbaum Associates. (Chapter 1).

Moussaid M., Helbing D. & Theraulaz G. (2011). How simple rules determine pedestrian behavior and crowd disasters. In the proceedings of the National Academy of Sciences of the United States of America, 108(17), 1-5.

Warren, W. H. (2006). The dynamics of perception and action. Psychological Review, 113(2), 358-89.

Warren, W.H. & Fajen, B.R. (2008) Behavioral dynamics of visually-guided locomotion. In A. Fuchs & V. Jirsa (Eds.), Coordination: Neural, behavioral, and social dynamics. Heidelberg: Springer.

Sebastian Burghardt, Bergische Universität Wuppertal,Wuppertal GERMANYArmin Seyfried, Bergische Universität Wuppertal,Wuppertal GERMANYWolfram Klingsch, Bergische Universität Wuppertal,Wuppertal GERMANY

Oneconsequenceofurbanizationisthegrowingcomplexityofbuildings.Tallskyscrapersandundergroundrailwaystationsarecommonbuildingsinlargecities,thereforstairshavetobeconsidered as part of the egress routes.

Thefundamentaldiagramisthebasicrelationtocharacterizetransportpropertiesoftrafficsystems.Forstairsthereisnoconsensuswhicharethemostimportantfactorsinfluencingthisrelation.Somehandbooksdistinguishbetweenup-anddownwardsmotion[1,3,4],othersfocusontheslopeofstair[2].Ofcourse,theflow-densityrelationfordifferentgeometriesisimportant,buttheavailabilityofmicroscopicanalysisalsoofferstheopportunitytotakeadeeperlookatinfluencesofmeasurementmethodandselectedmeasurementarea.Thusfurtheranalysislikespatialandtemporaldevelopmentofthebasicquantitiesvelocity,densityandflowcouldbeconsidered.Uptonow,theseinfluencesarenotscrutinizedandthefundamentaldiagramforstairs,eventheshapeofthefunction,hasn’tbeenunderstood.E.G.itisassumed,thatcongestioninfrontofstairsappearsduetolowercapacitiesofstairsincomparisontohorizontalroutes.Withmeasurementsprovidinghighresolutionintimeandspacewe found indications that the transition from the plane to the stair is responsible for the congestions andnotthesmallercapacitiesofstairs.Aclarificationofthisquestioncouldimprovetheconductofstairs and thus the design of emergency routes.

In this contribution an overview about the fundamental diagram for stairs is given. First we discuss discrepanciesoffundamentaldiagramsofwell-knownhandbooksforplanningofpedestrianfacilitiesandevacuationrouteslikePredtechenskiiandMilinskii[1],NelsonandMowrer[2],Fruin[3],andWeidmann[4].Totestthecorrespondencetorealmeasurements,wecollectpublishedmeasurementsavailableinliterature[6-11].Inthesecondpartwederiveafundamentaldiagramforstairsdownwardsbasedonownexperimentsandprecisetrajectories.Tocheckwhethertheresultsofexperimentsperformed under laboratory conditions are comparable with characteristics of motion of everyday situations,wepresentacomparisonwithresultsofafieldstudycarriedoutatthesamestaircase[5].Furthermorethiscontributionshowsamethodtogaintopographicalinformationofdensity,velocity,andspecificflowstructurestogetamicroscopicinsightintopedestriandynamicsonstairs.

Fundamental diagram of stairs: Critical review and topographical measurements of density and flow

REFERENCES

[1] V. M. Predtechenskii and A. I. Milinskii, Planning for Foot Traffic Flow in Buildings. Amerind Publishing, New Dehli, 1978, translation of: Proekttirovanie Zhdanii s Uchetom Organizatsii Dvizheniya Lyuddskikh Potokov, Stroiizdat Publishers, Moscow, 1969.

[2] H. E. Nelson and F. W. Mowrer, “Emergency Movement,” in SFPE Handbook of Fire ProtectionEngineering, 3rd ed., P. J. DiNenno, Ed. Quincy MA: National Fire Protection Association, 2002,ch. 14, pp. 367–380.

[3] J. J. Fruin, Pedestrian Planning and Design. Elevator World, New York, 1971

[4] U. Weidmann, “Transporttechnik der Fussgänger,” Institut für Verkehrsplanung,Transporttechnik,Strassen- und Eisenbahnbau, ETH Zürich, Tech. Rep. Schriftenreihe des IVT Nr. 90, 1993, zweite,ergänzte Auflage.

[5] S. Burghardt, A. Seyfried, and W. Klingsch, “Improving Egress Design through Measurement andCorrect Interpretation of the Fundamental Diagram for Stairs,” in Developments in Road Transportation, M. Panda and U. Chattaraj, Eds., Proceedings of the International Conference on Developments in Road Transportation, October 8-12, 2010, NIT Rourkela, Odisha, India. Macmillian Publishes India Ltd, 2010, pp. 181–187.

[6] X. Chen, J. Ye, and N. Jian, “Relationships and characteristics of pedestrian traffic flow in confinedpassageways,” Transportation Research Record: Journal of the Transportation Research Board, vol.2198, pp. 32–40, 2010, dOI: 10.3141/2198-05.

[7] H. Frantzich, “A model for performance-based design of escape routes,” Department of Fire SafetyEngineering, Lund Institute of Technology, Tech. Rep. 1011, 1994.

[8] Frantzich, H., “Study of movement on stairs during evacuation using video analysing techniques,”Department of Fire Safety Engineering, Lund Institute of Technology, Tech. Rep., 1996.

[9] M. Müller, “Fundamentaldiagramme von Personenströmen auf Tribünentreppen - Vergleich unterschiedlicher Steigungen am Beispiel der ESPRIT arena,” Bachelor Thesis, Bergische Universität Wuppertal, 2010.

[10] D. B. N. Seer, S. Bauer and M. Ray, “Estimating Pedestrian Movement Characteristics for Crowd Control at Public Transport Facilities,” in 11th International IEEE Conference on Intelligent Transport Systems, Oct 2008.

[11] J. Ye, X. Chen, C. Yang, and J. Wu, “Walking behavior and pedestrian flow characteristics for different types of walking facilities,” Transportation Research Record: Journal of the Transportation Research Board, vol. 2048, pp. 43–51, 2008, dOI: 10.3141/2048-06.

A data-driven model of pedestrian following and emergent crowd behavior

Kevin Rio, Brown University, Providence RI USAStéphane Bonneaud, Brown University, Providence RI USAWilliam H. Warren, Brown University, Providence RI USA

Canthecollectivebehaviorofhumancrowdsbeexplainedasanemergentpropertyoflocalpedestrianinteractions?Toaddressthisquestion,wehaveadoptedabottom-upapproachmotivatedbycognitivescienceandgroundedinexperimentaldata.Weutilizethebehavioraldynamicsframework(Warren2006),whichintegratesaninformation-basedapproachtoperception(Gibson1979)withadynamicalsystemsapproachtoaction(Kugler&Turvey1987;Kelso1995).Afullunderstandingofbehaviorinthisframeworkconsistsinspecifyinghowinformationabouttheenvironmentispickedupbytheagentandusedtocontrolaction(acontrollaw),andalow-dimensionaldescriptionoftheglobalbehaviorthatarisesasaresult(abehavioralstrategy).Weuseexperimentaldataonhumanlocomotiontoinformandtesthypothesesabouttheseprocesses,andtogeneratemodelsofpedestrianbehavior(Fajen&Warren2003;2007).

Thispaperappliesourapproachtothecaseofpedestrianfollowing.Onepersonfollowinganotherisacommonbehaviorineverydaylife,andmayprovideakeylinktothedynamicsofcrowdbehavior.Successfulfollowersmustcontroltheirspeedtostaybehindtheleader,andcontroltheirheadingtostayoncoursewiththeleader.Herewefocusonspeedcontrol.Ourgoalistomodelthebehavioralstrategyandthevisualcontrollawgoverningfollowingindyads,andthenusethismodeltosimulatethe emergent behavior of small crowds.

InExperiment1(withChrisRheaandJonCohen),weinvestigatedthebehavioralstrategygoverningfollowing.Wecollecteddatafrompairsofpedestrians,whowalkedonastraightpathina12x12mroomwhiletheirheadpositionsandorientationswererecordedusinganinertial/ultrasonictrackingsystem(IS-900,60Hz).The‘leader’(anexperimenter)walked3stepsataconstantspeed,thenspedup,sloweddown,orremainedatthesamespeedforseveralsteps(3,4,or5),andfinallyreturnedtotheoriginalspeed.The‘follower’(aparticipant)began1or3mbehindtheleaderandwasinstructedtofollowhimataconstantdistance.Wemodeledthetimeseriesofthefollower’saccelerationusingfourcandidatesderivedfromthedrivingliterature:(1)aspeedmodel(Chandler,Herman,&Montroll1958),basedonmatchingtheleader’sspeed;(2)adistancemodel(Pipes1953;Herman,Montroll,Potts,&Rothery1959),basedonmaintainingaconstantdistancebehindtheleader;(3)aratiomodelofspeedoverdistance(Gazis,Herman,&Rothery1961),and(4)aweightedlinearcombinationofspeedanddistance(Helly1959;Andersen&Saur2007).WeusedRMSEandPearson’srtoquantifygoodnessoffit.Theresultsindicatethatthespeedmodelperformedsignificantlybetterthanthedistancemodel,andthemorecomplicatedratioandlinearcombinationmodelsdidnotfurtherimproveperformance.Weconcludethatasimplespeed-matchingstrategyissufficienttoaccountforpedestrian following.

InExperiment2,ourgoalwastounderstandthevisualinformationthatisusedtocontrolfollowing.Thetwomostlikelysourcesofinformation(Rushton&Wann1999)arevisualangle,theanglethattheleadersubtendsatthefollower’seye,andbinoculardisparity,thedifferencebetweentheimageoftheleaderinthefollower’sleftandrighteye.Weusedvirtualrealitytodissociatethesevariables(Tarr&Warren2002),andindependentlymanipulatedthevisualangleandbinoculardisparityofavirtual‘leader.’Participantsfollowedavirtualtargetpole(1.65mheight,0.5mdiameter)viewedinastereoscopichead-mounteddisplay(SR-80A,63°Hx53°V)whiletheywalkedina12x12mroom.Headpositionandorientationweretracked(IS-900,60Hz)andusedtoupdatethedisplay.Achangeintargetspeedovera3sintervalwasspecifiedby(a)achangeinvisualangle,producedbyexpandingorshrinkingthepole,(b)achangeindisparity,producedbyincreasingordecreasingthepole’ssimulateddistancefromthefollower,or(c)achangeinbothvisualangleanddisparity.Theseconditionswerefullycrossed,sothatthechangescouldbecongruentorincongruent.Participants’meanspeedduringthemanipulationshowedthattheyreliedentirelyonthechangeinvisualangle(p<.001)andwerenotaffectedbythechangeindisparity(p>.05).Weconcludethatchangeintheleader’svisualangleissufficienttocontrolspeedinone-dimensionalfollowing.

Wecancombinetheseresultstoderiveacontrollawthatimplementsthebehavioralspeed-matchingstrategy.Ineffect,followersacceleratetonullchangesinthevisualangleoftheleader.Thisismorecognitivelyplausiblethanthespeed-matchingstrategy,becausethereisevidencethathumanagentsdonotaccuratelyperceiveobjectspeed(Rushton&Duke2009).

Armedwiththiscontrollawforpedestriandyads,canweuseittoscaleuptocrowds?InExperiment3,ourgoalwastoapplythecontrollawforone-dimensionalfollowingtostudysmallcrowds.Wecollecteddatafromfourpedestrianssteeringtowardacommongoal.Participantsbeganinasquareconfigurationofvariablesize(0.5,1.0,1.5,or2.5msides).Afterstartingtowalk,theyweredirectedtowardoneofthreegoalslocated8maway.Inapreviousstudy(Bonneaud,Rio,Chevaillier,&Warren2011),wefoundthatthesepedestrianscoordinatedtheirbehaviorbyadoptingacommonspeed.Wehypothesizedthatthespeedthatemergesisdrivenbypedestriansinthebackfollowingthoseinfront.UsingthecontrollawderivedfromExperiments1and2,wepredictedthetime-seriesofthefollower’saccelerationasafunctionoftheleader’svisualangle,andusedittosimulategroupbehavior.Thus,Experiment3demonstrateshowasimplecontrollawgoverninglocalpedestrianinteractionsmaycontribute to the emergent behavior of crowds.

Insum,wehaveusedthebehavioraldynamicsframeworktodevelopacognitively-plausible,experimentally-groundedmodelofpedestrianfollowing,whichhelpstoexplainthecollectivebehaviorofsmallcrowds.Thisapowerfulandunderutilizedapproachthatcanenhancepedestrianmodelingandinformfutureworkoncrowddynamics.

REFERENCES

Andersen, G. J., and Sauer, C. W. (2007). Optical information for car following: The driving by visual angle (DVA) model. Human Factors: The Journal of the Human Factors and Ergonomics Society, 49(5), 878-896.

Bonneaud, S., Rio, K., Chevaillier, P., and Warren, W. H. (2011). Accounting for patterns of collective behavior in crowd locomotor dynamics for realistic simulations. In Proceedings of the 2011 International Conference on Computer Animation and Social Agents.

Chandler, R. E., Herman, R., and Montroll, E. W. (1958). Traffic dynamics: Studies in car following. Operations Research, 6(2), 165-184.

Fajen, B. R., and Warren, W. H. (2003). Behavioral dynamics of steering, obstable avoidance, and route selection. Journal of Experimental Psychology: Human Perception and Performance, 29(2), 343-362.

Fajen, B. R., and Warren, W. H. (2007). Behavioral dynamics of intercepting a moving target. Experimental Brain Research, 180(2), 303-19.

Gazis, D. C., Herman, R., and Rothery, R. W. (1961). Nonlinear follow-the-leader models of traffic flow. Operations Research, 9(4), 545-567.

Gibson, J. (1986). The Ecological Approach to Visual Perception. New York, NY: Psychology Press.

Helly, W. (1959). Simulation of bottlenecks in single lane traffic flow. In Proceedings of the Symposium on Theory of Traffic Flow., Research Laboratories, General Motors (pp. 207–238). New York, NY: Elsevier.

Herman, R., Montroll, E. W., Potts, R. B., Rothery, R. W., Herman, R., Montroll, E. W., and Potts, R. B. (1959). Traffic dynamics: Analysis of stability in car following. Operations Research, 7(1), 86-106.

Kelso, S. (1995). Dynamic Patterns: The Self-Organization of Brain and Behavior (Complex Adaptive Systems). Cambridge, MA: The MIT Press.

Kugler, P. N., and Turvey, M. T. (1987). Information, Natural Law, and the Self-Assembly of Rhythmic Movement: Resources for Ecological Psychology. Hillsdale, NJ: Lawrence Erlbaum Associates, Inc.

Pipes, L. A. (1953). An operational analysis of traffic dynamics. Journal of Applied Physics, 24(3), 274.

Rushton, S. K., and Wann, J. P. (1999). Weighted combination of size and disparity: A computational model for timing a ball catch. Nature Neuroscience, 2(2), 186-90.

Rushton, S. K., and Duke, P. A. (2009). Observers cannot accurately estimate the speed of an approaching object in flight. Vision Research, 49(15), 1919-28.

Tarr, M. J., and Warren, W. H. (2002). Virtual reality in behavioral neuroscience and beyond. Nature Neuroscience, 5, 1089-92.

Warren, W. H. (2006). The dynamics of perception and action. Psychological Review, 113(2), 358-89.

Comparison of different calibration techniques on simulated data

Christian Rudloff, AIT Austrian Institute of Technology, Wien AUSTRIAThomas Matyus, AIT Austrian Institute of Technology, Wien AUSTRIAStefan Seer, AIT Austrian Institute of Technology, Wien AUSTRIA

Thesocialforcemodel[1]hasdevelopedintooneofthemainapproachesformodelingandsimulatingpedestrianmovementonamicroscopiclevel.Themodelisdefinedbydescribingtheaccelerationofanindividualpedestrianbasedonthesumofdifferentattractionandrepulsiveforces:whileanattractionforceacceleratespedestrianstowardstheirdesireddirectionandspeed,repulsionforcesallowevasionfrom other pedestrians and boundaries.

Manypapers(e.g.[2],[3],[4])discussmodeldevelopmentsandcalibratethesemodelsusuallyusingasinglecalibrationmethod(withtheexceptionof[4]wherethreemethodologiesarecompared)basedonpedestrianmovementobservations.However,itisnotknownwhatinfluencethechoiceofaspecificcalibrationtechniquehasonthequalityofthecalibratedmodelparameters.Inparticular,whencalibratingpedestrianmovementsimulationmodels(PMSM),onehastheproblemthatdatacollectedwithautomatedtrackingmethodsaswellasmanuallyannotationisnoisyduetoinaccuraciesofbothmethodologies.Additionally,mostdatacollectionmethodsdelivertrajectoriesbasedontheheadofpedestrianswhichincludesbodyswaycausedbythesidemovementofthepedestrian‘sheadduringeachstep.Thisintroducessignificantuncertaintiesasthepositionsandvelocitiesofthesurroundingpedestriansareusedtomodeltheaccelerationofapedestrian.Addingtothisisthefactthatthevelocityandaccelerationarecalculatedbydifferencingthenoisypositiondataleadingtoincreasederrorsforthosevariables.Whenusingamaximum-likelihoodestimationtoestimatethemodelparametersdirectlyfromtheobservedaccelerationdata,thismightleadtoerror-in-variablesproblems,resultinginabiastowardszerointheparameters(seee.g.[4]).

TogiveresearchersaguidelinewhencalibratingPMSM,thispapertestsseveralcalibrationtechniques.Therefore,wecreatedacalibrationdatasetwhichoriginatesfromsimulationsinsteadofobservationallowingfullcontrolovertherealparametersthatwereestimatedwiththedifferentmethodologies.Acorridorof2mwidthwithbidirectionalflowswaschosen.Overall50pedestriansweresimulated,25ineachdirection.Thetwogroupswereplaced10metersapartintoregular4x7grids.Apositionandgoalinoppositegridswereassignedrandomlytoeachpedestrianoutofthe28gridsquares.Allpedestrianswereassigneddesiredvelocitiesfromanormaldistribution(mean1.3m/s,variance0.2).Fivescenarioswerecreated.Theellipticalmodel2from[2]waschosenforbothrepulsionfromwallsandpedestrianstocreateourcalibrationdataset.Tocreatemorerealisticdata,anormallydistributederrortermwithmean0andahighvarianceof0.2wasaddedtoeachspatialpositionofthesimulateddata.

For our tests the parameters governing the interaction distance and the strength of the repulsive force betweenpedestrianswerecalibrated.Threedifferentcalibrationtechniqueswerecomparedonthe

differentdatasets.Thebasemethodwasamaximumlikelihoodestimationoftheparametersdirectlyderivedfromthecalculatedaccelerationsateachpointintime.Othermethodologiesfitthesimulatedtotheoriginaltrajectories.TwovariantswerecomparedbothusingthemeanEuclideandistancebetweensimulatedandoriginaltrajectoryasoptimizationcriteria:first,allpedestriansweresimulatedatthesametime,second,ineachsimulationrunonlyonepedestrianwassimulatedandallremainingpedestrianswerekeptontheiroriginaltrajectories.Inallcasesthedesiredvelocitiesofpedestrianswerecalculatedasthe90%quantileofthevelocitiesintheiroriginaltrajectories.TwodifferentMATLABoptimizationalgorithmswereusedforparameterestimation:thefirstfunctionappliedwasfminsearch,whichusesaNelder-Meadoptimizationalgorithm.Here,thestartingparametersfortheoptimizationwerealsovariedrangingfrom(a)theparametervaluesoriginallyusedtocreatethecalibrationdataset,(b)randomstartvalues.Thesecondmethodwasageneticalgorithm(GA)usingthefunctionga,startingwitharandompopulationof40.Asthemethodologydoesnotrelyon parameters found in real pedestrian behavior the parameter values were chosen using hints from literature and visual testing of the simulation.

Theresultsshowthatmaximumlikelihoodestimationmightbefeasibleifsmallerrorscanbeguaranteedintheinputdata.Otherwise,theerrorinvariableproblemwillresultinparameterswithabiastowardszero.Further,thecalibrationusingtheNelder-Meadalgorithmisstronglydependentonthestartingvaluesasthealgorithmtendstoonlyfindlocaloptima.TheGAdoesnotsufferfromastartingvalueproblemasitincorporatesrandomchangestotheparametervalues.However,duetothe random search the estimated parameters vary to a much larger extent if the search space cannot be restrictedatall.Comparedtotheparametervaluesof20and0.17forstrengthandinteractiondistanceoftheoriginaldataset,themeansfromtheGAestimationwere7.3and0.35withstandarddeviationsof9.8and0.35.UsingtheoptimalsolutionsoftheGAasstartingvaluesfortheNelder-Meadalgorithmimprovesthemeanvaluesto12.3and0.17.

It can be seen from our results that a larger strength parameter results in a smaller interaction distance. Thiscouldbeasignforanidentificationproblemforbothparametersleadingtoaregionofsimilarvaluesfortheobjectivefunction(seealso[2]).Thismightbedealtwithbyusingsimulationdatawithdifferentpedestriandensitiesassuggestedin[4].Thesamehappenswhentheerrorinthedataislarge.Thedifferencehereisthatthestrengthparametertendstobeestimatedmuchtoolargeandtheinteractiondistancetendstobetoosmall.Alikelyreasonisthatrandomerrorleadstoverystrongobservedaccelerations.Whenlookingatthequestionofsimulatingalloronlyonepedestrian,thepictureisnotquiteasclear.Forsmallornoerrorsintheinputdata,simulatingallpedestriansatoncegives more stable and better parameter estimates. For large errors the parameter sets are both showing large variation but simulating one pedestrian at a time leads to more stable estimates.

Inconclusion,thecombinationofageneticalgorithmfollowedbyanumericoptimizationstepseemstobemostpromising.However,atestwithmoresimulatedscenariosanddifferentdensitiesisneededtoverifytheidentifiabilityoftheparameters.Togetherwiththesepoints,infuturework,influencesofinherentproblemslikebodyswayneedtobeincludedintothesimulateddatatotestthemethodologiesfurther.Alsootheroptimizationcriterialikewalkingtimeinthesimulationversustherealwalkingtime should be tested and possibly combined with the trajectory distance.

REFERENCES

[1] D. Helbing and P. Molnar. Social force model for pedestrian dynamics, Physical Review E, 51 (5), 1995, pp. 4282-4286.

[2] A. Johansson, D. Helbing and K. Shukla. Specification of a Microscopic Pedestrian Model by Evolutionary Adjustment to Video Tracking Data, Advances in Complex Systems, 10 (2), 2007, pp. 271-288.

[3] S. Hoogendoorn and W. Daamen. Microscopic Parameter Identification of Pedestrian Models and Implications for Pedestrian Flow Modeling, Transportation Research Record, 1982, 2006, pp. 57-64.

[4] C. Rudloff, T. Matyus, D. Bauer and S. Seer. Can walking behavior be predicted? An analysis of the calibration and fit of pedestrian models, to appear in Transportation Research Records.

Individual assessment of conflict prediction and group dynamic behavior as main driver for pedestrian dynamics

Michael Schultz, Technische Universität Dresden, Dresden GERMANYLars Rößger, Technische Universität Dresden, Dresden GERMANY

Modelsforpedestriandynamicscopewithdifferentaspectsofbehaviorrelatedtohumanmovements.Generally,suchmodelscanbeassignedtothreedifferentlevelsofmovementcharacteristics:operational,tacticalandstrategicbehavior.Thebasicmicroscopicmodels(e.g.socialforce,cellularautomat,ordiscretechoice)particularlyfocusontheoperationallevel(spatialexclusionordistanceanddirectionrelatedrepulsion).Especially,thefavorablesocialforceapproach,whichstatesattractionandrepulsionforcesbetweenthehumanbeings,turnsoutasagoodanalogytoreproducesubstantialself-organizationeffects.Severalmodelmodificationsandextensionsofthesocialforcemodelhavebeenrecentlydevelopedbythescientificcommunity.Onecannoticethatsustainableconcepts(e.g.discretization,floorfields)willbetransferredbetweenthedifferentmodelapproaches,andthatthemodelswillconvergeasanevolutionaryconsequence.

Insteadofrefiningthestandardideasofmovementmodelingourcurrentresearchprojectslayemphasis on the pedestrian perception and cope with both the dynamic group behavior and the individualevaluationofpotentialconflictsituations,andhence,ratherfocusonthetacticallevelofmovementbehaviour.Whereasmodelsmentionedaboveprimaryconsideroperationalaspects(spatialexclusionordistanceanddirectionrelatedrepulsion),theconsiderationofpsychophysicalconcepts,likeourapproachofregardingtheperceivedtime-to-conflict,overcomestheideaofunspecificdirectedrepulsionforcesandderivesspecificmovementdecisionwithrespecttotheindividualevaluationofpotentialconflictsituations.Theideaofconsideringtheindividualhumanperceptiontocopewith enhanced patterns of movement behavior is from our point of view the next challenge for the upcomingresearchtasks.Aninterdisciplinaryresearchapproachincludingtheresearchareasoftrafficsciences,sociology,mathematics,physicsandpsychologywillensurethatourproposedmethodsfollows a common agreement of all parties involved.

Thedatarecordingtovalidateourdevelopedfundamentalmethodologyconsistsofthreesources:(a)acquisitionofdatainthefield,(b)isolatedtestenvironmentsforstepwiseparameterevaluation,and(c) complex test scenario to cope with the individual and complex interactions.

(a)Forthedataacquisitioninthefield,werecordedthemovementbehavioroftheparticipantsoftheGermanProtestantKirchentagatDresden(1.-5.June2011with120.000fulltimeparticipantsandapprox.50.000guests)andusethisdataasasolidbaseforthegroupconstellationandbehavior.Asourdatapointsout,therearesignificantdifferencesinthedensity-speed-relation(fundamentaldiagram)regardingtheconstellationofgroups.Heterogeneousgroupsconsistsofindependentpedestrianspossess a homogenous density and each pedestrian has a high flexibility to change the speed and the directionofmotion.Theeffectof“clustereddensity”(localdensityspots)increaseswiththeamountof

groups,theirmobility,andwiththegroupsize.Thesedensityspotssignificantlychangetheindividualspeedcharacteristicandthecorrespondingavoidingbehavior.Obviously,eachgroupmemberplanstheindividualandgroupmovementsrespectingtheoverallgroupbenefitandconsidertheParetoruletoensureacommonmovementagreement.Duetothefactthatthisplanningprocedureoftenbasedonnon-verbalcommunication,thisprocedurewillfailwithincreasingdensity.Groupswithaclearleader and follower structure as well as experienced groups contain of members with a comparable hierarchystatuswillefficientlysolvethis“synchronized”movementtaskevenundercrowedsituations.Weobservedthismovementpatternsoverawidearea,whichleadinahighervarianceofspeedinsidethepedestrianflowcomparedtotheexpectedstandarddistributions.Highlyagilegroupsarebenefitfromthedensityclustersbyefficientlyuseofthecorrespondingfreeaccessiblespace.Soitseemedthatamixtureoftwodifferentflowsexitsinsidethepedestrianstream.Thesestructuresarestableinanenvironmentupto1.5-2persons/m²,butabovethisdensitytacticalmovementsarebarelymanageableandthetwoseparatedflowsarecombinedtoone.Westronglysuppose,thattheongoingdataevaluationwillprovidefurtherfactstoapprovethesefindings.

(b)Theisolatedtestenvironmentfocusesontheindividualassessmentofconflictprediction,whereourmodelconcentratesonatime-to-conflictmethod.Modelparametersarethedifferentperceptionareas,thepointofview(angle)dependedperceptionofspeedandpathdeviations.Acomparisonamongthe movement components speed and path deviation shows that the path deviation can easily be estimatedifthetrafficisoncomingbutthecorrespondingspeedismuchhardertopredict.Conversely,inthecaseofanorthogonallyapproximation,thiseffectstranspose.Withregardtotheenvironmentalperception,theindividualcharacteristics(e.g.physicsorassertiveness),andthegroupconstellationweinvestigateddifferentconflictavoidancestrategies.Apreliminaryevaluationoftherecordspointsoutacleardominationofjustafewavoidancepatterns.Consideringthispromisingresults,wewillsetupcontinuativetestenvironmentstoprovethesefindings.

(c)Alsotocopewiththepsychologicalinteractionsbetweenhumanbeingstheresultsofevacuationexperimentsareincludedinourpedestrianmodelapproach.Inevacuationexperiments,weobservedagaintheexistenceofheterogeneousgroupsintermsofmovementparameters.Theresultssuggestthat-eveninsituationswhicharecharacterizedbyhighdegreesofuncertaintyandurgency-individualsmorelikelyusesocialcuesfromtheimmediatesurroundingsocialenvironmentfortheirindividualmovementdecisions.Obviously,suchsocialcuesreducetheuncertaintyincriticalsituationsandsupportdecisionmakingwhenowndecisionrulesarenotoronlylimitedavailable.Hence,pre-existinggroupaffiliationdeterminestacticaldecisionsonanindividuallevelandleadsto“clustereddensity”oncumulative(collective)level(fromamacroscopicpointofview).Furtherin-depthanalysesfocusingon personality traits and movement related decisions indicated that certain traits (such as extraversion) areassociatedtotheextentindividualsusesocialinformation.Thesepreliminaryfindingsofgroupdynamicsunderlinetheneedforconsideringcharacteristicsofdecisionmakingprocesses(withrespecttosocialinteractionsandonindividuallevel),andthus,tofocusonthetacticalcomponentwithinourcomprehensive pedestrian modeling approach.

Evening presentation


Thomas Brudermann, Universität Graz, Graz AUSTRIA Mass psychology revisited - Insights from social psychology, neuroscience and simulation

Thistalkrevisitstheclassicalfieldofmasspsychology(alsoreferredtoas“crowdpsychology”).Startingwithanoverviewonrelatedconceptsfromsocialpsychologyitdiscussesphenomena,wheremasspsychologyplaysavitalrole:Humanpanicsandstampedes,butalsostockmarketbubbles,fashiontrendsorpoliticalmovements.Althoughofdifferentnature,phenomenadrivenbymasspsychologyhaveacommonground:Individualdecisionmakingisreplacedbypsychologicalcontagion,andfactsarereplacedbyopinionsaboutfacts.Infurtherconsequencecontagioncascadesmightappear,causingirrationalandhardlypredictablecollectiveoutcomes.

Technologicalandmethodologicalprogressesinrecentyearsadvancetheunderstandingforsuchmassdynamics.Inparticular,neurosciencesontheonehandandagent-basedsimulationsontheotherhandareofgreatvalue:Whileneurosciencesallowforabetterunderstandingofhumanbehaviorontheindividuallevel,agent-basedsimulationsclosetheanalyticgapbetweenindividualdecisionsandcollectiveoutcomes.Acombinationofinsightsfrombothfieldshencecontributestoabetterunderstandingofmassdynamics.

Keynote

Thursday, 7 June - 09.00

Paul I. Weinmann, ICTS Europe, Amsterdam NETHERLANDS Behavior assessment of individuals in crowds - a view from the perspective of aviation and airport security

Internationalterrorismhaslongbeenasignificantthreatfactorforthetravelindustry,pervading many aspects of mass transportation worldwide and generating dynamic developments in transportation security.

Civilaviationinparticularremainsoneofthemostattractivetargetsforterrorism,andtheupcomingOlympicGamesinLondonraisethethreatlevelsignificantly.Accordingtointelligenceandevidencepublicizedrecentlyinthemedia,terroristsarerelentlessindevisingnewmodioperandiandimprovedtechniquesforconcealingweaponsandexplosivematerials,aimedatoutwittingsecuritytechnologiesandbypassingsecurityprocessesatairports.Theirtargetsareairplanes,crowdsofpeopleatairports,andotherairportfacilities.Investmentinthecreationofmulti-tierairportsecuritysystems,inresearch,andinthedevelopmentofsophisticatedsecuritytechnologieshasimprovedcivilaviationsecurity,especiallysincetheeventsof9/11.Nevertheless,asterroristmethodsbecomemoresophisticated,technologyaloneislimitedinitsabilitytodetectnovelmeansofattack.Also,thetechnologicalapproachtosecurityengendersmoreandmoresecuritylayers,increasingthealreadycontestedhassletopassengers and other airport users.

Thesecuritymethodologybrieflyillustratedinthepresentlectureaimsfirstandforemostatdetectingtheterroristintentofindividuals,ratherthanweaponsandexplosivematerial.Duringthewholecourseofinformation-gatheringonpotentialtargets,planning,andupuntiltheverymomentofattack,terroristswilltrytodisguisetheirrealidentitiesandintentsbyusingfalseIDsandotherdocumentation,pretendingtohaveinnocentaims(i.e.,beingregularpassengers),andbycamouflagingtheirdangeroustools.Suchattemptstodeceivesecuritystaffresult,however,inpatternsofappearanceandbehaviorandinotherindicationsthatcouldsuggestterroristintent.Trainedsecuritystafffamiliarwithsuchindications,whicharealsoknownas“suspicioussigns”(asopposedtoincriminating,“hard”evidence)areabletodetectandevaluatethesesignsinrealtime,byapplyingmoderate,polite,privacy-mindedmeanssuchasverbalinteraction,interview,andoccasionallyinspectionofsupportingdocumentationanddata.Theprocesswillthenresult–ifdeemednecessary–inmorediscerningandappropriatesecuritymeasures,thuscomplementing,orevencompensatingfor,theshortcomingsofsecurity technologies.

Thisrisk-basedandpro-activemethodologyaimsatoptimizingthesecurityprocessbyidentifyingpotentiallyhigh-riskindividualsandfocusingeffortsonthem.Suchsecuritylayerneedstobeimplementedwithintheframeworkofstandardprocessesandsecurityproceduresatairports,anditcanalsobeintegratedintoothercustomaryairportactivities.

Oral presentations


A.13 Rahul Jobanputra, University of Cape Town, Cape Town SOUTH AFRICA The development and calibration of an agent-based microsimulation model to simulate vehicle-pedestrian interaction

A.14 Heiko Aydt, Nanyang Technological University, Singapore SINGAPORE Symbiotic simulation for egress optimisation in smart buildings

A.15 Christoph Dobler, Eidgenössische Technische Hochschule, Zürich SWITZERLAND Integration of a microscopic force-based 2D pedestrian simulation into a framework for large-scale transport systems simulation

B.13 Helmut Schrom-Feiertag, AIT Austrian Institute of Technology, Wien AUSTRIA Simulation of handicapped people finding their way through transport infrastructures

B.14 Maria Davidich, Siemens AG, München GERMANY Calibration and validation of pedestrian simulations against real live scenarios based on video data: Example of a German railway station

B.15 Toshihiro Osaragi, Tokyo Institute of Technology, Tokyo JAPAN Simulation model of evacuation behavior following a large-scale earthquake that takes into account various attributes of residents and transient occupants

C.13 Julien Pettré, INRIA, Rennes FRANCE Velocity-based models for crowd simulation

C.14 Dorine Duives, Delft University of Technology, Delft NETHERLANDS Analysis of pedestrian self-organizing crowd movements at a music festival

C.15 Günter Bärwolff, Technische Universität Berlin, Berlin GERMANY Methods for modeling and simulation of multi-destination pedestrian crowds

The development and calibration of an agent-based microsimulation model to simulate vehicle-pedestrian interaction

Rahul Jobanputra, University of Cape Town, Cape Town SOUTH AFRICAMarianne Vanderschuren, University of Cape Town, Cape Town SOUTH AFRICA

Withtheincreasesincomputerprocessingpowerandadvancesinprogrammingskills,anarrayoftransportationandurbanplanningcomputermodelsarenowavailabletotheprofession.Theyareextensively used in developed nations to model complex transport scenarios and interactions. Models varyfrommacroscopic,whichfocusonthesystemasawholeandahigheraggregationlevel,tomorecomplexmicroscopicmodels,whichallowthesimulationofindividualroadusersandbehaviourtoobtainmorerealisticrepresentationsatalocal/streetlevel.

Themodels,tovaryingdegrees,giveresearchersandpractitionerstheabilitytoanalysetheeffectivenessofinterventionsonadisaggregatedlevel,asindividualvehiclesand/orpedestriansaresimulatedindetailastheymovethroughtheroadnetworkwiththegoalofreachingtheirdestinationbythemostcosteffectiveorshortestroute(Harney,2002).Despitethis,researchersindicatethat,todate,themajorityofmicroscopictrafficmodeldevelopmentandsimulationhasessentiallyfocusedontheanalysisoftransportationefficiency,suchassignalisedintersections,arterialnetworks,freewaycorridorsandcrowdevacuationordynamics(Cunto,2008).

Theadvancesintechnologyhave,however,ledtothedevelopmentofcommercialmicroscopicmodels,whichprovidetheopportunitytosimulatevehicleandpedestrianinteractionundervaryinginfrastructureconditions.Thesetypesofmodelsaregainingrecognitionasmethodsforassessingurbanplanningstrategiesaswellasmeasuringandpredictingsafetyand,offerthepotentialforproactive safety analysis for all road users rather than just vehicles.

Modelssimulatetheinteractionofroadusersviaparametersettingsof,forinstance,compliancelevels(i.e.attitudetorisk),car-followingdistances,acceleration/decelerationspeeds,walkingspeeds,aggressionandawarenesslevels,etc.Thevaluesoftheseparametersareincorporatedinthesoftwareatdefaultlevelssetbythevendorsandaffectoutcomesofmodelledscenarioswhichmaynotreplicateactualroaduserbehaviourandthuscurrentsystemperformance.Calibrationofthesevaluesis,therefore,vitaltomatchobservedconditionsandtoproducerealisticoutputs.Directmeasurementofthesevaluesisverydifficultbecausemanyofthemrepresentsubtlefeaturesthatarehardtoisolateand,becauseoftheextensiveamountofdatacollectionrequiredatadisaggregatelevel.

Thecomplexityofmicrosimulationmodelsandthelargenumberofparametersthat,usually,requirespecificationmeansthatthenatureofcalibrationrequiredisacomplexanditerativeprocess.Becauseofthis,manyresearchershavereducedthenumberofparametersbasedonheuristicsandonatrialanderrorbasis.Additionally,theliteraturecitesseveralstudieswiththeprimarygoalofprovidingarobustcalibrationmethod.Earlyresearchworkfocussedonsearchalgorithmsforcalibrationbasedon

asinglecriterion.However,itissuggestedthatthisapproachfailstorecognisethattrafficisamulti-facetedentity(Duong,2011).Multi-criteriacalibrationhasbeenproposedasanimprovementtothisbyanumberofresearchers(forexampleToledoetal,2004;Balakrishnaetal,2007);Maetal,2007;Ciuffoetal,2008;HuangandSun,2009)butclearlydependsonrelativeweightingoffactors.Researchison-goingintomethodsofremovinganybiasingoffactorsusingthismethod.

ThispaperdetailstheevaluationofparametersforasimulationstudyofalocalarterialroadinCapeTown,SouthAfrica,usingacommerciallyavailablepackage,Paramics.Resultsarecomparedtovaluesobtainedusingdefaultsettingsandparameterscalibratedbyusingthemethodsdescribed.Thestudyindicatesthat,althoughtheAgent-basedversionofthissoftwareallowsflexiblemodellingofvehiclesandpedestrians,anddespitethecalibrationofthestudynetworkusingmethodsdescribedabove,itdoes not allow the replication of observed local road user behaviour by vehicles and pedestrians at crossings.Themodificationofseveralofthesebehaviouralparametersispossibleinthesoftwareviaaprogrammablemodule.Giventheconclusionofthestudy,theinvestigationintothepossibilityofmodifying appropriate parameters to replicate observed local road user behaviour is also presented.

REFERENCES

Balakrishna, R., Antoniou, C., Ben-Akiva, M., Koutsopoulos, H.N., and Wen, Y. 2007. Calibration of microscopic traffic simulation models. Transportation Research Record: No.1999, pp. 198–207.

Ciuffo, B.F., and Punzo, V., and Torrieri, V. 2008. Comparison of simulation-based and model-based calibrations of traffic-flow microsimulation models. Transportation Research Record, No.2088, pp. 36-44.

Cunto, F. 2008. Assessing Safety Performance of Transportation Systems using Microscopic Simulation. Thesis presented to the University of Waterloo, Canada.

Doung, D. 2011. Calibration and validation of VISSIM microscopic traffic simulation model parameters using Pareto Archived Dynamically Dimensioned Search. PhD Student Paper, University Avenue West, Waterloo, Ontario.

Dowling, R., Skabardonis, A., Halkias, J., McHale, G., Zammit, G. 2004. Guidelines for Calibration of Microsimulation Models. Journal of the Transportation Research Board, pp.1-10.

Harney, D. 2002. Pedestrian modelling: Current methods and future directions. Road & Transport Research Dec 2002.

Huang, W and Sun, J. 2009. A NSGA-II based parameter calibration algorithm for traffic microsimulation model. Proceedings of the IEEE Computer Society: 2009 International Conference on Measuring Technology and Mechatronics Automation.

Toledo, T., Ben-Akiva, M., Darda, D., Jha, M., Koutsopoulos, H. 2004. Calibration of Microscopic Simulation Models with Aggregate Data. Transportation Research Record, No. 1876, pp.10-19.

Symbiotic simulation for egress optimisation in smart buildings

Heiko Aydt, Nanyang Technological University, Singapore SINGAPOREMichael H. Lees, Nanyang Technological University, Singapore SINGAPOREStephen J. Turner, Nanyang Technological University, Singapore SINGAPOREWentong Cai, Nanyang Technological University, Singapore SINGAPORE

Increasing population and the trend of urbanisation results in more and more densely populated cities.Managingtheincreasingnumberofpeopleincitiesefficientlyhasbecomepriorityformanycitygovernments.Whileurbanisationhasmanyadvantagesitcomeswithitsownproblems.Most(ifnotall)largecitieshavetodealwithproblemsthataretypicalfortheirsizeanddensitysuchastrafficjams,forexample.Smarttrafficsystemscanhelptoalleviatetheproblemstosomeextent.Anotherconsequenceofincreasingpopulationdensityincitiesarehigh-floorbuildings(e.g.,officetowers,megashoppingmalls,high-riseresidentialbuildings)thatprovideshelter,officespace,entertainment,andrecreationforthousandsofpeople.Althoughbuildingshavetosatisfycertainsafetystandardsbydesign,forexamplebyfeaturingwell-displayedemergencyexitsandwell-locatedfireextinguishers,thelargenumberofoccupantsmaycausesignificantcongestionincaseofanecessaryevacuationduetoeventssuchasfire.

Technologycanhelptomakeoperationsmoreefficientandegressisnoexception.Infact,aswewillshowinthispaper,egressisagoodexamplewheretechnologycanmakeapositivedifference.Buildingsaretypicallyequippedwithfloorplansthatshowemergencyexitsandevacuationroutestothenearestexit.Whilesuchastaticguidancesystemcangivepeopleageneralideawheretheyarecurrentlylocatedandwherethenearestexitis,itcannotprovideguidancethatadaptsdynamicallytothecurrentsituation.Incaseofanemergency,asituationmaybecomeverydynamicforvariousreasonssuchasrapidspreadoffireandsmoke,congestedevacuationroutes,blockedexits,orevenpanic.Theabilitytocaptureinformationaboutasituationinreal-timerepresentsanimportantadvantagethatcanhelptomakeegressmoreefficient.Smartbuildingscanbeequippedwithvariousformsofsensors(e.g.,smokedetectors,heatdetectors,motiondetectors)andactuators(e.g.,electronicsign-boards)thatcanbeutilizedbyaguidancesystem.Thisguidancesystemcanusereal-timesensordatatodynamicallyadapttothecurrentsituationandprovideusefulinformation(e.g.,directiontothenearest/safestexit)totheevacuees by means of various actuators.

Giveninformationaboutthecurrentsituationinthebuilding,theguidancesystemcancomputethepreferredroutetoanexitfromanylocationinthebuilding.Thereareanumberofchallengesthatneedtobeconsideredandadequatelyaddressed.Oneimportantissue,thatwewilladdressinourwork,isconcernedwiththedynamicsofanemergencysituation.Weareconcernedwithdynamicsthatarisefromtherapidspreadoffireandsmokeaswellashumancrowdbehaviour.Inparticular,crowddynamicsmayleadtocongestionsandbottle-necksincertainpartsofthebuilding.Although,areactivezero-lookaheadguidancesystemcanprobablytakeintoconsiderationthenumberofpeopleinthevariouspartsofthebuildingaswellasthecurrentlocationofthefire(s)inordertodetermineanidealrouteinordertoavoidcongestion,itlackspredictivecapabilities.Thiscanbeaserious

disadvantageandmayevenexacerbatecertainproblems.Inthispaper,weproposeasimulation-basedguidancesysteminasmartbuildingenvironment.Thissystemmakesuseofhigh-fidelitysimulationsthatenableittopredicthowtheemergencyunfoldsinthenear-future.Forthispurpose,wewillutilizeideasfromthefieldofsymbioticsimulation[1,2].

Symbioticsimulationisaparadigminwhichaphysicalsystemandasimulationsystemarecloselycoupledbysensorsandactuators.Thisrelationshipisoftenmutuallybeneficial.Thesimulationsystembenefitsfromreal-timesensordatawhichmakesitpossibletoperformhigh-fidelitysimulationsofthephysicalsystem.Thephysicalsystem,ontheotherhand,benefitsfromtheoutcomeofwhat-ifanalysesconductedbythesimulationsystem.Thepurposeofsuchawhat-ifanalysisdependsontheapplication.Here,wedescribeasmartbuildingapplicationwheretheguidancesystemisbasedonasymbioticsimulationsystem.Therefore,inthisparticularapplication,thesymbioticsimulationsystemisconcernedwithadecisionmakingproblem:whichisthebestevacuationpathataparticulartimeinaspecificpartofthebuilding?Thesymbioticsimulationsystemiscapableofevaluatingalternativeroutes(e.g.,what-ifscenarios)bymeansofsimulationastheegresseventprogresses.Theabilityofasymbioticsimulationsystemtosimulatemanypossiblewhat-ifscenarios,enabletheguidancesystemto analyse various possible solutions and select the one which provides the best performance (in terms ofevacuationtimeorsafety,forexample).

Inthispaper,wedemonstratetheeffectivenessofthesymbioticsimulation-basedapproachthroughvariousexperiments.Theexperimentshighlightthesituationsinwhichasmartbuilding,equippedwithanactiveguidancesystem,canbemosteffectiveinhelpingpeopleescapefromabuilding.Theseexperimentsareperformedusinganagent-basedcrowdmodel.Forthesimulatedbuilding,weconsideratypicalmulti-storeyofficebuilding.Morespecifically,thebuildingmodelisadaptedfrompubliclyavailablefloorplansoftheWorldTradeCenterinLongBeach,California.

REFERENCES

[1] R. Fujimoto, D. Lunceford, E. Page, A. U. (editors), Grand challenges for modeling and simulation: Dagstuhl report, Tech. Rep. 350, Schloss Dagstuhl. Seminar No 02351 (August 2002).

[2] H. Aydt, S. J. Turner, W. Cai, M. Y. H. Low, Symbiotic simulation systems: An extended definition motivated by symbiosis in Biology, in: Proceedings of the 22nd Workshop on Principles of Advanced and Distributed Simulation, 2008, pp. 109–116.

Integration of a microscopic force-based 2D pedestrian simulation into a framework for large-scale transport systems simulation

Christoph Dobler, Eidgenössische Technische Hochschule, Zürich SWITZERLANDGregor Lämmel, Technische Universität Berlin, Berlin GERMANY

Transportsimulationisanimportantresearchtopicformanyyears.Ingeneraltherearetwodifferentareasoftransportsimulations.Ononehand,modelshavebeendevelopedforthesimulationoflarge-scalescenarioswithhundredsofthousandsorevenseveralmillionentities.Ontheotherhand,therearemodelsforsmallerscenarioswithsomehundredorafewthousandentities.Theformerclassofmodelsusuallydealswithvehiculartraffic,whereamicroscopicmodelingofthephysicsisnotneededanone,instead,candrawonsimplermodelsknownfromthefieldofdynamictrafficassignment.Thelatterclassofmodelsusuallydealswithpedestrians,wherethescenariosareoftenrelatedtoevacuationsituations. In those scenarios a microscopic modeling of the underlying physics seems to be necessary. Inrecentyears,theinterestinmulti-modalsimulationmodelshasincreasedsignificantly.Insuchmodels,varioustransportmodesaresimulatedsimultaneously,includingtheinteractionsbetweenagentsusingdifferentmodes.Typicalfieldsofapplicationare,forexample,studiesoncarsharingandpublic transport.

However,attemptstoimplementamulti-modalsimulationhavetosolvetheproblemthatthecomputationalefforttosimulatelarge-scalescenarioswithamicroscopicmodelisenormously.Toovercomethisproblem,wepresentanapproach,wherethelevelofdetailwithinamodelcanvary.Bydoingso,itis,forexample,possibletomodelinteractionsbetweencarsandpedestriansonlyatselectedareas.Tosimulatetheinteractionbetweenthetransportparticipants,amicroscopicmodelingisneeded.Thiscallsforamulti-agentsimulation,whereeverytransportparticipantisrepresentedbyasoftwareagent.Inthesimulation,agentsmakeindependentdecisions.Thiscanforexampleberouteanddestinationchoice(e.g.wheretogoshoppingandwhichroutetotaketotheshoppingmall).Theuse of agents in a transport simulation allows to model the human behavior in a realistic manner.

Thispaperintroducesacombinationofboth—macroscopicandmicroscopic—approaches,wherethevehiculartrafficissimulatedwithaso-calledqueuemodelwhilethepedestrianmovementissimulatedbyaforce-basedmodel.Theobviousscenariosforsuchanapproacharesituationswheretheagentsarearriving at a location by one mode of transport and then switching to another mode of transport.

Themainadvantageofthequeuemodelisitscomputationalefficiency,whichisabasalrequirementforlarge-scalesimulations.Inthismodel,thetransportsystemistransformedintoanetworkoflinksandnodes.Eachlink(streetsegment)isrepresentedasaFIFO(first0-infirst-out)queue.Everyagenthastoremainonthelinkforacertaintime(freeflowtraveltime).Eachlinkhasaspecificoutflowcapacity,which corresponds to the flow capacity of the associated street segment. If at any time the outflow capacityisusedup,nomoreagentscanleavethelink.Furthermore,eachlinkhasaspecificstoragecapacity.Ifitisusedupnomoreagentscanenterthelink.

Intheforce-based2Dsimulation,theagents’high-levelplanning(i.e.routeanddestinationchoice)isperformedonagraphrepresentingthetransportsystem,whilethelowlevelbehavior(i.e.physicalinteractionbetweentheparticipants)issimulatedwithaforce-basedmodel.Intheforce-basedmodelthe simulation entities are emitting repelling (other agents and obstacles) and attracting forces (goal locations).Theforce-basedmodelitselfisbasedonexistingwellestablishedapproaches.

Achallengingtaskistomodeltheswitchfromonemodetoanother.Thisisparticularcomplicatedwhentheinvolvedmodesaresimulatedwithdifferentsimulationmodels—i.e.switchfromthequeuemodeltotheforce-basedmodelorviceversa.Anexampleisanagentwhoarrivesbycarinaparkinglot,simulatedbythequeuemodel,andthenswitchestoatrue2Dsimulationmodel.Thereasonisthatdifferentmodelsaresimulatingondifferentphysicalresolutionsbutneverthelessinfluencingeachother. Theproposedapproachisdevelopedasanextension—whichisbasedon[1]—totheMATSimframework.MATSimstandsforMulti-AgentTransportSimulationandiswidelyusedinthetransportsimulationcommunity.Themainfieldofapplicationisthesimulationoflarge-scalevehiculartraffic.Balmer[2]givesadetaileddescriptionoftheframework,itscapabilitiesanditsstructure.MATSim‘sapplicationtoalarge-scaleSwitzerlandscenario(over6millionagentssimulatedonahighresolutionnetworkwith1millionlinks)ispresentedbyMeisteretal.[3].MATSimisalsoappliedtootherscenarioslikelarge-scalepedestrianevacuationsimulationsorthesimulationofairtransport.However,sofarallapplicationsarebasedonthequeuemodel.

Theintroducedmulti-modalmodelistestedonahypotheticalscenario.Inthescenario,theagentsarriveatametrostationorparkingareanexttoashoppingmall.Afterleavingthemetrostation,theagentshavetocrossastreetbeforetheyenterthemall,goshoppingandreturnbacktothemetrostation.Oncetheagentsareagainatthemetrostation,theygetonthenexttrain.Agentsarrivingattheparkingareacanwalkdirectlytothemall.However,theaccessroadtotheparkingareacrossesthefootpathbetweenthemetrostationandtheshoppingmall.Inthisscenariotherearethreedifferentmodesoftransport.First,therearetrainsservingthemetrostation.Second,therearepedestriansmovingfromthemetrostationtotheshoppingmall.Thepedestriansaresimulatedbytheforce-based2Dmodel.Andthirdthereistrafficontheaccessroadtotheparkingareanexttotheshoppingmall.

Thenoveltyinthispaperisthecombinationofsimulationmodelsofdifferentscales.Theproposedapproachgivestheopportunitytosimulatelarge-scalescenarios,whilestayinghighlyresolvedwhereneeded and being more aggregated where possible.

REFERENCES

[1] G. Lämmel (2011) Coupling force-based and graph-based models for multi-agent wayfinding in complex environments, Working Paper, 11-20, VSP, TU Berlin, Berlin.

[2] Balmer, M. (2007) Travel demand modeling for multi-agent traffic simulations: Algorithms and systems, Ph.D. Thesis, ETH Zurich, Zurich, May 2007.

[3] Meister, K., M. Balmer, F. Ciari, A. Horni, M. Rieser, R. A. Waraich and K. W. Axhausen (2010) Large-scale agent-based travel demand optimization applied to Switzerland, including mode choice, paper presented at the 12th World Conference on Transportation Research, Lisbon, July 2010.

Simulation of handicapped people finding their way through transport infrastructures

Helmut Schrom-Feiertag, AIT Austrian Institute of Technology, Wien AUSTRIAThomas Matyus, AIT Austrian Institute of Technology, Wien AUSTRIA

Publictransportinfrastructuresmustaccommodatelargenumberofpassengerswhiletakingintoaccountspecificneedsofcertaingroupssuchaselderlyorhandicappedpeople.Theirdesignneedstofacilitatewayfindingforpeopleunfamiliarwiththeinfrastructure.Moreover,thelayoutmustprovideforsufficientspacesuchthatdelaysduetocongestionareminimized.

Thusthedesignprocessisanontrivialtaskwhichnowadaysissupportedbypedestriansimulationstopredictpassengerflowsalreadyintheplanningstage.Currentmodelsdonotintegrateadetaileden-routeroute-choicemodelthatimplementsorientationornavigationbehaviorbutratheruseshortest-path routing only applicable for people familiar with the infrastructure.

However,differentgroupshavedifferentneedsfortheirroutesandtheirorientationdependingontheirsensoryandmobilityimpairments.Atthefirstvisitofaninfrastructuredependingonthesignagealreadyfindingtheelevatorcanbecomeachallengingtask,especiallyforpeoplehavingreducedreceptioncapabilities.Theseindividualsneedmuchlongerfortheirorientation,andpotentiallytemporarilyblockpassagefortheothers,ifthedesignoftheinfrastructuredoesnotsupporttheirneedssufficiently.

Consequentlythispaperdiscussesaresearcheffortputintoenhancingexistingsimulationmodelsbyincludingmodelsforthemotionandorientationbehaviorofmobilityimpairedpassengerslikeindividualswithprams,wheelchairusers,individualswithsensoryimpedimentsandpeoplebeingunfamiliar with the infrastructure.

Heretogatherinformationongroupspecificbehaviouracomprehensivefieldexperimentsweremadewith47people.Eightgroupsofpeoplewereidentifiedwhomaydemonstratemobilitypatternsclearlydistinguishablefromthegeneralpopulation:70+age,peoplewithpram,visuallyimpaired,blind,wheelchairusers,mobilityimpaired,hearingimpairedanddeaf.Themainresearchfocuswastoinvestigatedifferencesinwalkingspeed,patternsingaininginformationfromtheenvironment,orientation and navigation.

TheexperimenttookplaceatamajortransithubinVienna,the“BahnhofNord”.Foreachpersonintheexperimentoneoutoftwodifferentscenarioswithtypicalusagepatternsanddifferentlevelsofcomplexitywasselected.Thescenarioscontainspecifictaskssuchasbuyingaticketanddrinkforthejourney,locatingtimetableinformation,orusingtherestroomsinthestation.

Empiricalmethodswerecombinedinordertogainrelevantqualitativeandquantitativedataonpedestrianbehaviour.Adetaileddescriptioncanbefoundin[1].Togathertheexperimentdataanobservationtechniquetypicallycalled“shadowing”wasused,seee.g.[2].Herebyanobserverfollows the observed individual recording the major directional changes using a special application onatabletPC.Additionaltheobservedpersoncarriedavoicerecordertodocumentthewayfindingprocess(“thinkingaloud”).Withthecombinationofthesetwomethodsitispossibletoobservewhichguidanceinformationwasrecognizedatwhichpositionsinsidetheinfrastructure.

Theanalysisoftheexperimentdata,showninthispaper,haverevealedontheonehandsurprisingsimilaritiesbetweenverydifferentgroups,ontheotherhandextremedeviationswithinasinglegroup,indicatingthatinsomecasesdeterminantsotherthandisabilityplayamoresignificantrolefortraveltimeandnavigationbehaviour.Thespatialanalysisoftrajectorydatarevealedthemainroutes,deviationsfromtheseroutesandclusterswherepeoplemainlystop.Theanalysisofthethinkingalouddata related to the trajectory helped to identify elements of the guidance system that respondents used to navigate and pointed out typical areas for orientation in the infrastructure.

Themaincontributionofthispaperistoproposeasimulationmodelthatrepresentsthespecificgroupsandbehaviours.Anagentbasedapproachisusedallowingthecharacteristicsofindividualpedestrianstobeassignedandvariedasrequired.Thehumanmotiononanoperationallevelismodelledbasedonasocialforcemodel[3].Asafirstapproximationthebasicsocialforceequationsareusedformodellingpedestrianandwheelchairmovementsvaryingtwoparameters:1)Thedesiredspeedoftheagentonwhichtheattractiveforceand2)thehorizontalbodysizeonwhichtherepulsiveforces are depending.

Onthetacticallevelthecognitionofguidancesystemsismodelledandmakesitpossibletosimulateagentnavigationthroughanunknowninfrastructureusingthepresentsignage.Noroutinggraphhastobedefinedinadvance,onlytheinformationobtainedfromthesignagemodelledin3Disnecessary.Themainroutingstrategyissearchingrandomlyintheabsenceofanyinformation.Informationisobtainedfromsignageelements(signs,monitors)inacertainareainfrontofthesignageelement.Theprocessofsearchingforinformationconsistsoftwophases,firstasignageelementisidentifiedandwalkedtowards,subsequentlyatasmallerdistanceandasuitableangletowardsthesignageelementtheinformation provided is absorbed.

Intheexampleofthetrainstation“ViennaNorth”themonitorsaretheprimarilysignagetogaininformation.Afterknowingthedepartureinformationlookingforthewaytothetrainplatformisthenextstep.Thetrainplatformsareindicatedusingasignconveyingtheplatformnumber.Thevisibilitydepends on the agent’s vision capabilities or a high crowdedness respectively (especially for wheelchair driverswithalowerpointofviewcanbemodelled).A3Denvironmentincluding3Dmodelsforthe simulated agents gives the opportunity to calculate possibly seen sections from one viewpoint realistically.

Finallyinordertodemonstratethefunctionalityofthewayfindingalgorithmdifferentscenariosarediscussedinthispaper.Startingattheelevatorintothemainhallofthe”ViennaNorth”trainstationawheelchairdriverhastogetthetrainto”Stockerau”.Allscenariosarebasedonthesametasksinpredefinedorder:Firstgotothetoilet,thenbuytheticket,lookforamonitortogetthedepartureinformation,gotothesupermarketandfinally,usetheelevatoruptotheplatformwherethetrainisdeparting.Thethreescenariosdifferinthelevelofcrowdednessandtheagent’svision:1)emptyhall,fullvision,2)emptyhall,halfvisionand3)crowdedhall,fullvision.Eachscenarioshoweddifferent

REFERENCES

[1] V. Egger, H. Schrom-Feiertag, G. Telepak, and L. Ehrenstrasser, “Creating a Richer Data Source for 3D Pedestrian Flow Simulations in Public Transport,” Measuring Behaviour 2010.

[2] Bauer, D., Brändle, N., Seer, S., Ray, M., and Kitazawa, K., “Measurement of pedestrian movements - a comparative study on various existing systems,” in Pedestrian behaviour: Models, data collection and applications, ed. H. Timmermans, Emerald Group Publishing, 2009, pp. 301-319.

[3] D. Helbing and P. Molnar, “Social Force Model for Pedestrian Dynamics,” Physical Review E, vol. 51, pp. 4282–4286, 1995.

routescomparedtotheshortestrouteasexpectedfor1)andsimilarlongerroutesbutdifferentdeviationsfor2)and3).Especiallyforwheelchairuserscrowdednesshasasignificantimpact.Crowdsoftenoccludethesignageorformobstaclesleadingtoahigherrateofmanoeuvre,longerroutesandtravel discomfort.

Theproposedmulti-agentbasedsimulationmodelfacilitatesanagenttofindautonomouslyitswaythroughabuildingbasedonsignageinformationonlyandmakesitpossibletoevaluatethevisibilityoftheguidancesystemfordifferentgroups.Itcanrevealareaswithleaksofguidanceinformationforpeopleunfamiliarwiththeinfrastructure,especiallyforelderlyandhandicappedpeoplewithreducedreception capabilities.

Calibration and validation of pedestrian simulations against real live scenarios based on video data: Example of a German railway station

Maria Davidich, Siemens AG, München GERMANYGerta Köster, Technische Universität München, München GERMANY

Whenplanningabuilding,or,forexample,organizingaspecialevent,itisofvitalimportancetominimizerisksduetocriticalsituations.Itwouldbeveryusefulifonecouldpredictsuchacriticalsituationforacrowd.Herepedestrianstreamsimulationsareveryhelpful.Theysimulatebehaviourofacrowdandthusallowplayingthroughdifferentcriticalscenariossuchasevacuations.However,apedestrian stream simulation can only be useful if it is capable of reproducing real situations.

Ifonewantstouseacertainsimulatorforaconcreteapplication,thebestwayistotestthesimulatorbeforehand by comparing simulation data to data that was extracted from live observations and stored. In most cases measurements from live data are not available and simulations are calibrated basedonknownliteraturedataorlaboratoryexperiments.ThemostwidespreadapproachistotakeaGaussiandistributionforthefree-flowvelocityandWeidmann’sfundamentaldiagramfortheflux-densityrelation[1].Here,threequestionsoccur:First,aretheseassumptionsrelevantforanyreallifescenarios?Second,isitenoughtohavethisinformationtocalibrateasimulationagainstareallifescenario?Andifnot,whatotherparametersarenecessarytodoit?Lastbutnotleast,howitispossibletodeterminetheaccuracyofasimulation?Thisworkisstrivestoanswerthesethreequestions.

Inthisworkwedemonstrateamethodologyforadjustmentofpedestriansimulationstolivescenarios.WedemonstratethisontheexampleofaGermanrailwaystation.Weshowthemethodologystepbystep,startingwithgatheringdatawithvideocamerasinstalledatamajorGermanrailwaystation,thenconductingdataanalysis,thencalibratingthesimulationtoolagainstthedatafromthelivescenarioandfinallyvalidatingthesimulationresults.

Westartwithextractingdatafromvideos:SeveralcameraswereplacedonahighceilingonamajorGermanrailwaystation.Thetrajectoriesofpedestrianswereextractedmanuallyfrom1,5minutesvideosusingatoolthatallowsto„click“positionsonthevideo.Asaresultforeachvideoweobtainedaround400trajectoriesofindividualpedestriansintimeandspace.Alltrajectorieswithintheexaminedareawereanalyzed.Weeliminatedonlythetrajectoriesthatweredistortedbydistanceorlargelyobscuredfromthecameraforvelocitydistributionflow-densityanalysis.However,thesetrajectorieswereusedforsource-targetsstatisticsandtoobtainascheduleofappearanceanddisappearanceofpedestriansfromasource.Thedetailedanalysisofvelocitiesandflow-densitydistributionisconductedexclusivelyonthevisibleandundistortedpartsofthetrajectoriestokeepmeasurement errors small.

Weextractedthefollowinginformationfromthevideostoadaptoursimulationtooltothereallifescenario:

• Thetopologyoftheareaofinterestintwodimensions.Thisisdirectinputdata.• Thepositionsofsourcesandtargetswithinascenario,thatis,thelocationswherepeoplecome fromandtowheretheygoto.Thisisagaindirectinputdata.• Statisticalinformationonthedistributionoftrajectoriesbetweensourcesandtargets.This guarantiesthatpedestrianswillwalkintherightdirections.• Ascheduleofpedestrianappearancesanddisappearances.• Thescenariospecificdistributionsoffree-flowvelocities.• Measureddatafromwhichthedensity-flowrelationship(fundamentaldiagram)validforthe currentscenariocanbederived.Ourobservationisthatitisveryimportanttousethe measuredfree-flowvelocitydistributionthatiscorrectforaparticularscenario.

Wewouldliketostressthatinsomescenariosadditionalinput,liketheaveragesizeofpedestrians,maybenecessaryoratleastbeneficial.

Theanalysisofthedatafromourlivescenariorevealedsignificantdifferencestoknownliteraturedata[1],whichunderlinestheimportanceofscenariospecificmeasurements.Pedestriansappearslowerthanausuallyacceptedvalue1,34m/s,andtheyslowdownfasterthanitissuggestedinaWeidmanndiagram.

Wetestedthesuccessofourapproachwiththehelpofapedestriansteamsimulationobtainedfromabenchmarkmodelbasedonacellularautomaton.Clearly,noexactmatchofeveryindividualtrajectoryofeverypedestriancanbeexpected,atleastsincewehaveastatisticoftrajectoriesandeverystartofsimulationwillprovideaslightlydifferentpicture.Therefore,noindividualtrajectoriescanbecompared.Weneedanaggregatedquantityinstead.Wepickthedensityofthecrowdasitevolveswithtimeinanareaofobservation.Thedensitycannotonlybemeasuredquiteeasilyinbothcases,butisalsoofimmediateinterest,becausedensitiesaboveacertainthresholdwouldbeanindicatorforimpending danger.

Wecomparedensityatdifferentpartsofthescenario.Thedensityevolutioninsimulationreproducestheobservedonavideowell,apartfromthefactthatitsometimesslightlyoverestimatesit.However,ifthesimulationisusedasaplanninghelptoavoidcriticalsituationsorason-linetoolforthepredictionof crowd congestion a slight overestimate can be tolerated whereas underestimate of density is inacceptable.

Thus,weproposedanapproachofadjustmentsimulationagainstareallifescenario.Theproposedmethodwastestedonacomplexreallifescenario--amajorGermanrailwaystationduringrushhours.Thesuccessoftheproposedapproachwasshownbycomparisonofsimulationwithvideos.Thecriticalparameters for adjustments are discussed.

Simulation model of evacuation behavior following a large-scale earthquake that takes into account various attributes of residents and transient occupants

Toshihiro Osaragi, Tokyo Institute of Technology, Tokyo JAPANTakayuki Morisawa, Tokyo Institute of Technology, Tokyo JAPAN

INTRODUCTIONInordertoconstructasimulationmodelforevacuation-effortsfollowingasevereearthquake,weconsidernotonlythespatiotemporaldistributionofoccupantsremaininginthecity,butalsooftransientoccupants,(i.e.,peoplewhoarewalkingorotherwiseintheprocessofusingtransportationinthecity).Themodeltakesintoconsiderationthebehavioralcharacteristicsofavarietyofpopulations,some of whom are constantly moving and some of which remain in a single location within the city. Wehaveattemptedtoaccountfortheinfluencebothpeoplewhosemovementpathsstaywithinthecityandthosereturninghome(e.g.,fromanoffice)onthenumberandspatialdistributionofrefugees.Theseaspectshavenotbeenaddressedinpreviousstudies.

STUDYAREAANDANALYTICALDATATheregionemployedinthisresearchwastheSetagayaWardofTokyo.Setagayahasareasthataredenselycrowdedwithwoodenstructures,andthusareveryvulnerabletohousingcollapseandurbanconflagrationintheeventofadevastatingearthquake.Inaddition,manyworkersandstudentsliveinSetagayaandcommutewithintheTokyoregion(i.e.,theywouldbecountedastransientoccupants).Locationdatafortherailwaypassengers,automobileusers,andpedestrianswerepredictedemployingthe results of previous research together with the detailed data available on individual attributes based onperson-tripsurveydata(PTdata).

MODELINGPROPERTYDEMAGEInamodeldescribingbuilding-collapse,theprobabilityofcollapsecanbeestimatedforeachbuildingunitbasedonitsyearoferectionoritsconstructiontype.Weuseddataonbuildingheightsandstreetwidthstomakeaforecastofstreetblockagecausedbybuildingcollapse.Inamodeldescribingfirespread,amodeloftheTokyoFireDept.wasemployed,whichcanaccountfordifferentmodesoffirespreadanddifferentstructuralbuildingtypes.

MODELINGHUMANBEHAVIORSThetypicalseriesofhumanbehaviorswhenanearthquakestrikeswerecategorizedandmodeled.Wemodeledthebehaviorofoccupantsbeingathome,fromtheperspectiveofinitialreactions,searchingforroutes,methodsofevacuation(directevacuationversustwo-stageevacuation),walkinginacrowd(walkingspeed),andactivitieswhilewaiting.Thebehavioroftransientoccupantsandoccupantsinfacilitiesotherthantheirownresidencesarealsomodeledbytakingintoaccounttheircharacteristics.

EXECUTIONOFTHESIMULATIONMODELIn order to examine the importance of accounting for the presence of transient occupants and persons returninghome,weanalyzedthevariationsinpeopleinputintothesimulationfor3specificcases:(1)thepopulationconsistingonlyofoccupantsremaininginthearea,(2)thatpopulationplustransientoccupants,and(3)thatpopulationplustransientoccupantsandpersonsreturninghome.

Incase(2),thetime-basedchangesintransientoccupantsstronglyaffectthenumberofrefugees,specifically,includingthemincreasesthenumberofrefugees,andincase(3),thenumberofrefugeesisgreatlyincreasedbytheinclusionofpersonsreturninghomeinthedaytime,whentherewouldbemanypersonsreturninghome.Examiningtheresultsbyagegroup,thereisanincreaseinchildrefugees(5–12years)inthemorningandevening,buttheelderlypopulation(65ormoreyears)showedlittlevariation.Incontrast,refugeesinthe13-to-64-yearagegroupgrewbyabout20%duringthe08:00-to-17:00timeframe.Morespecifically,thepresenceoftransientoccupantsandpersonsreturninghome,whichhaspreviouslybeenunaccountedfor,maycauseanapproximately20%increaseinthenumberofrefugeesestimatedforthepeakhoursperiod.Thisisanamountthatcannotbeignored.

Next,weconsiderthenumberofrefugeesinevacuationareas.Incases(2)and(3),thereappearedalarge influx of refugee railway passengers at the evacuation areas around locations with high densities ofrailwaylines.Previousinvestigationsregardingthenumbersandlocationsofevacuationareashaveusuallybeenbasedonthenumbersoflocalresidents,butinanurbanareawithahighlydevelopedtransportationnetwork,itseemsessentialfortheplanningofevacuationareastoaccountforthepresenceoftransientpopulationsonrailwaysandothertrafficarteries,especiallyduringthemorningand evening rush hours.

Thesimulationwasexecuted,accountingforalloccupants(whetherinbuildings,transient,orontheirwayhome),andtherisksduringevacuationtoopenareaswereevaluated.Thecasesoftwo-stageevacuationanddirectevacuationwerecompared.Sinceresidentsathomemakeupamajorityoftherefugeesatnight,theriskisgreatlyreducedbyorganizingatwo-stageevacuation.Duringtheday,however,manyoftherefugees,specificallyoccupantswhoarenothomeresidents,wouldnotcomplywithatwo-stageevacuation.

SUMMARYANDCONCLUSIONAsimulationmodelwasconstructedtodescribehumanactions(waiting,returninghome,andevacuating)afteranearthquake,takingintoaccountpropertydamage.Anumericalsimulationofpeople’sreactionsafteradisastrousearthquakeinSetagayaWardofTokyowasemployedtoverifytheimportanceofaccountingforthepresenceoftransientoccupantsandthatofpersonsreturninghome,andtheimportanceofcombiningsuchamodelwithonedescribingpropertydamage.Ouranalysishasrevealedthatthepresenceoftransientoccupantsandpersonsreturninghome,whohavenotformerlybeenincludedindisasterpreventionplanning,mayaddabout20%tothenumbersofrefugeesduringmorningandeveningpeakperiods,aquantitythatcannotbeignored.Two-stageevacuationswerealsoevaluatedfromtheviewpointofrisksduringevacuationtoopenareas.Thisindicatedlargevariationsinthenumbersandattributeprofilesofrefugees,dependingontherefugelocationandthetimingofthe disaster.

Velocity-based models for crowd simulation

Julien Pettré, INRIA, Rennes FRANCE

Severalfieldspaidalotofattentiontocrowdsimulationtheselastyears.Thisparticularinterestresultedinnumerouscrowdsimulationmodelsofvariousnatures.Thecomputergraphicscommunityhas participated to this proliferation to answer its needs ranging from the production of movies to the developmentofvideogames.Actually,themotivationsofthegraphicscommunityarenotveryfarfrom the one of transports science:

-Realisticsimulationisexpected.Thecomputergraphicsismoreinterestedinvisualrealismofresults,and often refer to believability to avoid strictly evaluating simulation models and comparing them to reality.Nevertheless,realityremainsacrucialreferenceandthegraphicscommunitystartscomparingtheir simulation results with real measurements.

-Emergentcollectivephenomenonisstronglydesired.TheseminalworkfromReynolds[4]onflockingbehavioremergedfromthegraphicscommunity.ItwasparticularlyinterestingbecauseReynolds’modelhasabilitytosimulatetheemergenceofcollectivebehaviorswithvisuallyimpressiveresults.Again,graphicscommunityisinterestedinsimulatingtheemergenceofrealisticcollectivebehavior.Simulationmodelsdevelopedinthegraphicsfieldstartbeingaccuratelyevaluatedonthispoint,includingbycomparingsimulationswithmeasurementsofrealcollectivebehaviors.Thecommunityalsonowcontributestheeffortincollectingdata.

-Realismisexpectedbothatthemacroscopicandthemicroscopiclevelsbecausebothscalescontributetothevisualaspectofresults:aclose-upviewofsimulatedcrowdsisenabled.Agent-basedmodelsarethemostfittingthisneedforrealismatdifferentscales.

Themainobjectiveofthispaperistopresentseveralmodelsproposedinthegraphicsfieldwhichcanbeofinterestfortransportscience,and,whereastheyareoftenformulatedoncommonbasis,todetailtheirsubtledifferences.

Agent-basedmodelisoftenidentifiedasasingleclassofapproachtocrowdsimulationbyitself.Actually,severalsubcategoriesofapproachesshouldbedistinguishedandcompared.Especially,wedistinguishatleasttwosubcategories:position-basedmodelandvelocity-basedmodel.Thesetwodifferenttypesofmodelscanbedistinguishedasfollows.Wedenoteq_ithestateoftheagenti.Aposition-basedmodelformulatesinteractionsbetweentwoagentsiandjasafunctionf(q_i,q_j)oftheirrespectivestate,whereasavelocitybasedmodelisalsobasedonthefirsttimederivativeofthesestatesf(q_i,q‘_i,q_j,q‘_j).

Atypicalexampleofposition-basedmodelisthewellknownsocialforcemodel[5]:theinteractionsbetween agents are formulated as some repulsive forces which are modeled as functions of the relativepositionsofagents.Socialforcesmodel,or,moregenerallyforce-basedmodelsreceivedalotofattentionandwerewidelystudied.Theirdrawbacksarealsowellknown.Inparticular,theyfailreproducingrealisticindividuallocomotiontrajectories,because,atthelocalscale,theyarebasedonunrealisticfoundations:indeed,apedestrianisnotrepulsedbyanyonethatwouldbecloseenoughtohim,andhemayavoidsomeoneatsomedistancewhenrequired.

Actually,ourresearchteamgathersexpertiseinhumanmotionscienceandcomputeranimationandfocusedonhowdohumandoperformcollisionavoidancemaneuvers.Westudiedpairwisecollisionavoidance(1vs.1particpant)andcheckedvarioushypothesisprovidedbymodels.Wealsostudiedmorecomplexsituationswithgroupsofparticipantswalkinginvariousgeometries.Byusingaccurateoptoelectronicsmotioncapturesystemstostudytheirmotion,wecoulddeducethebasicmotionlawsasusedbyhumanstoperformacollisionfreelocomotion.Wededucedcollisionavoidancemodels,and,moreprecisely,chronologicallyproposedtheParispredictivemodel[1],asimilarmodelwithsimplifiedformulationandcalibrationonexperimentalmotiondata[2],andasyntheticvision-basedmodel[3].

Thetwofirstmodelsshareidenticalfoundations.Thevelocityspaceisexploredanddecomposedintotwomajorcomponents:theadmissibleandnon-admissiblevelocityspacesthatrespectivelycorrespondtovelocitiesthatenablecollisionfreemotionornot.Thesetwomodelshowevercomputethesetwocomponentsindifferentways:

-theParismodel[1]exploresadelimitedtime-windowandsearchforanoptimaladmissiblevelocity.Basedonsomesimplificationoftheexpressionofrelativemotions,themodelisabletomergetogethermultipleinteractionsinanefficientway.

-thesecondone[2]isentirelybasedontheexpressionofrelativevelocityvectorsandisnotlimitedtoatimewindow,butittakesintoaccountuncertaintyinmotionperception(i.e.,intheexpressionofvelocityvectors)toreproducerealistichumanbehaviors.Especially,inordertoreproducethetimingofaninteraction,themodelsuccessivelyreporducestheobservationphase,thereactionphaseandtheregulation phase of an interaction.

Thetwofirstmodel[1]and[2]workunderthefollowinghypothesis:interactionarecombinedtogether(whenapedestrianfacesseveralinteractions)bysortingthemaccordingtotime-to-collisionandmergingthen-firstsbyexploringthecommonsolutionspace.Thevision-basedmodel[3]exploresaradicallynewpath.Thismodelisbasedonthefollowingassumptions:

-Humanbehaveaccordingtosimplisticperception/actionloops-Humanimplicitlycombinemultipleinteractionsbytheirvision(asa2Dspacewherealltheinformation is projected)-Humanprocessthiscomplexstimulusbyselectingmosturgentsituations

Weproposetocontinuedetailingthedifferencebetweenthesemodelsalongtheextendedversionofthepaper,addcomparisonwithothermodelsfromthegraphicsfield(e.g.,RVO[6]),aswellasprovidingmoregeneralperspectivesonthefutureofvelocity-basedmodels.

REFERENCES

[1] S. Paris, J. Pettré and S. Donikian. Pedestrian Reactive Navigation for Crowd Simulation: a Predictive Approach. Computer Graphics Forum, Vol. 26, N° 3, Eurographics’07 Conference Proceedings, p. 665-674, 2007.

[2] Julien Pettré, Jan Ondřej, Anne-Hélène Olivier, Armel Crétual and Stéphane Donikian. Experiment-based Modeling, Simulation and Validation of Interactions between Virtual Walkers. Proceedings of the 2009 ACM SIGGRAPH/Eurographics Symposium on Computer Animation (SCA‘09), 2009.

[3] Jan Ondřej, Julien Pettré, Anne-Hélène Olivier and Stéphane Donikian. A Synthetic-Vision-Based Steering Approach for Crowd Simulation. SIGGRAPH ‚10: ACM SIGGRAPH 2010 Papers, 2010.

[4] Craig W. Reynolds. Flocks, herds and schools: A distributed behavioral model. SIGGRAPH Comput. Graph. 21, 4 (August 1987), 25-34, 1987.

[5] D. Helbing and P. Molnár. Social force model for pedestrian dynamics. Physical Review E 51, 4282-4286, 1995.

[6] Stephen J. Guy, Jatin Chhugani, Sean Curtis, Pradeep Dubey, Ming C. Lin, and Dinesh Manocha, „PLEdestrians: A least-effort approach to crowd simulation,“ in Proceedings of the 2010 ACM SIGGRAPH/Eurographics Symposium on Computer Animation, Madrid, Spain, July 2-4, 2010, pp. 119-128.

Analysis of pedestrian self-organizing crowd movements at a music festival

Dorine Duives, Delft University of Technology, Delft NETHERLANDSDr. W. Daamen, Delft University of Technology, Delft NETHERLANDSProf. Dr. S.P. Hoogendoorn, Delft University of Technology, Delft NETHERLANDS

INTRODUCTIONMajorpedestriancrowdmovementshaveproventobevolatileinthepast(e.g.Cambodiawaterfestival2010,DuisburgLoveParade2010).Simulationbymeansofpedestrianmodelsisusedtopredictpossiblehazardouslocationsandtimeintervalsatmajorpedestriancrowdeventsandassuchcanhelppreventpedestriancrowdaccidents.Thecontributionofthisresearchistobuildonthecurrenttheoreticalknowledgethatqualitativelyandquantitativelydescribesthetransitioneffectsbetweentheself-organisingregimeswithinpedestriancrowdmovements.UsingatoaUAVattachedcamera,footageofpedestriancrowdmovementswasrecordedatamajormusicfestivalintheNetherlands.Thevisiblecrowdmovementsareanalyzedusingacompletelynewdevelopedtrackingtool,resultinginbetterdefinitionofthetransitionbetweenself-organisationregimes.

BACKGROUNDANDLITERATUREREVIEWSeveralscientistshavetriedtodescribepedestriancrowdbehaviour,withafocusonpublicspaceplanninganddesign.Eventuallyvariousspeed-densitycurveswereextractedbymeansofexperiments(Fruin1987),(Predtetschenski1971),observationofnaturalbehaviour(Helbing2005,Alghadi2002)andtheanalysisofdisasters.Alloftheaboveconcludedthathighercrowddensitiesreduceindividualwalkingspeeds.Fangetal.(Fang2003)combinedtheworkofFruin,PredtechenskiiandMilinskiiandsubtractedanupperandlowerboundonthedensity-velocitygraph,whichclearlyshowsthelackofconformitybetweentheseattemptstodefinepedestriancrowdbehaviourbasedonsolelyvelocityandspeed.

Additionally,severalformsofself-organisationhavebeendescribedinliterature,e.g.bi-directionallaneandstripeformation,zipper-effectatbottlenecksstop&gowaves(Helbing.2005),turbulentbehaviourandpressureshockwavesinstill-standingcrowds(Helbing2001).However,theexactrelationshipbetweenthepedestrianflowcharacteristics(especiallyunderhighdensities)andtheformsofself-organisingbehaviourispoorlyunderstood.Thiscouldbecausedbythelackofstableandlongtimecoverageofhigh-densitypedestriancrowdmovements.

Manytoolshavebeenusedtomakepedestriantrackingwithrespecttotrafficflowexperimentseasier,suchascolouredcaps,whitet-shirts,contrastingwallsandfloorsandpredefinedlimitedareasofmovement.However,thesetoolscanonlybeusedinstableresearchenvironments.

Inthisresearchthreemainquestionswillbeanswered.Firstofall,canourproposedobservationtechniquebeusedtorecordandanalysepedestriancrowdmovements?Secondly,cananewtoolbedevelopedthatcandetectandtrackmultiplepedestriansfromareasonablybigheights(50–100m)

wherenopreliminarymeasurestofacilitateidentificationareused?Andlastofall,canthetransitionbetweenself-organisingregimesbespecifiedbasedonthemacroscopicflowcharacteristics(speed,density and intensity)?

METHODOLOGYWithanewdatacollectiontool-anoctacopter,equippedwithalightweighthigh-speedhigh-definitioncamerawithaF3.4541.435mmLeicalens–wearenowabletofilmoutdoorpedestriancrowdmovementswithoutinterferingwiththenaturalmovementofthepedestriancrowd.UnlikemostCCTV,anairbornecamerarecordsatop-viewimagefromaheightof50metersorhigher,whichensureslimitedocclusionofpedestriansandalargerecordingarea.Fig.1.bdisplaystheimagesrecordedthisyearataDutchmusicfestivalnamedLowlands.Thetotalareacoveredbythecamerais500by350metersandshotunderanangleofapproximately49degrees.Onaverage250–800pedestrians are visible in every frame. Vibrationandinterlacingandocclusionarepresentintheimagesequences.Furthermoreduetotherelatively small dimensions of a singular pedestrian there is only limited information available on each pedestrian.ThereforeacombinationofthelenscalibrationandorthorectificationproceduredevelopedbytheTUDelft,backgroundtrackingopticalflowcalculationsandacolourmixturemodelwasusedtotrackeachpedestrianinthecrowd.Dependingonthedensityoftheflowthismethodcanaccuratelyaccountforthe80–90percentofthepedestrianmovementswithintheimage.

Afterthedatacollection,theformsofself-organisationwithinthecrowdareinvestigatedbasedontheformation(physicallocation)ofvisitorswithinthecrowd,themacroscopicflowvariablesspeed,density and intensity and the microscopic variables spacing and angular deviations are used in the investigation.Opticalflowcalculationsareusedasanestimationofspeed.Densityisdeterminedbasedonthenumberofpedestrianspersurfacearea(5x5meters).Countingthenumberofpedestrianscrossingpredeterminedbordersintwoseparatedirections,intensityintwodirectionsiscalculated.Spacingbetweenpedestriansiscalculatedusingthedistancebetweenanindividualandotherpedestrianswithinthevisionfield.Thelastvariablerecordstheshiftbetweenthecurrentdirectionandthe following movement direction.

RESULTSTherecordedpedestrianmovementsaretranslatedintotimeseriesofeachoftheabovedescribedvariables.Additionallywewereabletodeductindividualtrajectoriesfromtherecordings,whichcreatedthepossibilitytoalsodeduceindividualtrafficcharacteristics(e.g.speed,angulardeviation,headwaysandanOD-matrix).Besidestheflowcharacteristicsalsotherecordedself-organisationregimesweremappedwithinthesequences.Threepredominantformsofself-organizationarevisiblewithinthesequences,namelylane&stripeformation,stop&gowaves,andthelackofbothregimes(dominantflowin1direction).Becausethefoundself-organisationformshaveverydistinctmacroscopicflowcharacteristics,thesecharacteristicscanbeusedasanindicationoftheappearanceoftheself-organisationforms.Thetransitionintothelastregimegivesanindicationoftheflowcharacteristicsthatdetermineself-organisation‘break-down’..Itwasfoundthatparticularlydensity,crowdspeedandcrowdformationsplayanimportantroleduringbreak-down.Whetherthelackofself-organisationissolelycausedbyeitheroneorbyacombinationofthesethreeisstillbeinginvestigated.Itisexpectedthatthefinalversionofthispaperwillexplaintheirroleinthebreak-downprocess,oratleastattempttodoso.

REFERENCES

Algadhi. S.A.H., H. Mahmassani, R. Herman. „A speed-concentration relation for bidirectional crowd movements,.” In Pedestrian and Evacuation Dynamics, door S. Sharma Schreckenberg. M, 3-20. Springer, 2002.

Fang, Z., S.M. Lo and J.A. Lu. „On the relationship between crowd density and movement velocity.” Fire Safety Journal, 2003: Vol. 38, pp. 271 - 283.

Fruin, J. Pedestrian - planning and design. Mobile, Alabama: Elevator World, 1987.

Helbing, D., P. Molnár, I. Farkas, K. Bolay. „Self-Organizing pedestrian movement.” Environment and Planning B: Planning and Design, 2001: Vol. 28, pp. 361 - 383.

Helbing., Buzua, Johansson. „Self-Organized Pedestrian Crowd Dynamics: Experiments, Simulations and Design Solutions.” Transportation Science, 2005: Vol. 39, pp. 1-24.

Hoogendoorn, S.P., H.J. van Zuylen, M. Schreuder, B. Gorte, G. Vosselman. „Microscopic Traffic Data Collection by Remote Sensing.” Transportation Research Board 2003 Annual Meeting, 2003: pp. 1 - 11.

Predtetschenski, W.M., Milinski, A.I. Personenströme in Gebäuden. Berlin: Staatsverslag der Deutschen Demokratischen Republik, 1971.

Methods for modeling and simulation of multi-destination pedestrian crowds

Günter Bärwolff, Technische Universität Berlin, Berlin GERMANYFrank Huth, Technische Universität Berlin, Berlin GERMANYGregor Lämmel, Technische Universität Berlin, Berlin GERMANYKai Nagel, Technische Universität Berlin, Berlin GERMANYMatthias Plaue, Technische Universität Berlin, Berlin GERMANYHartmut Schwandt, Technische Universität Berlin, Berlin GERMANY

INTRODUCTIONRapidgrowthinthevolumeofpublictransportandtheneedforitsreasonable,efficientplanninghas made the description and modeling of transport and pedestrian behaviors an important research topicinthelasttwentyyears.Inthestudyofpedestrianbehavior,evacuationscenarios(inwhichpedestriansalltargetadefinitedestination)andmulti-agentsystems(inwhichpedestriansaretreatedasheterogeneousindividuals)haveattractedmuchattentionastwospecificproblems.Comparativelylittleattentionhasbeenpaidtotheproblemofpedestriancrowdbehaviorsingeometries with multiple destinations: each of the possibly many pedestrians moves to one out of a numberofdestinations.Theobjectiveofthepresentstudyistoinvestigatepedestrianbehaviorinsuchacontext.Thecentralproblemisthemodelingofcrossingpedestrianstreams.Inviewofadesirablepracticalrelevance,realistic,i.e.rathercomplexgeometriesarestudiedinthiscontext.

EXPERIMENTSInordertoobtainreliableempiricaldataofmulti-directional,intersectingpedestrianflowsfortheevaluationofdifferentsimulationmodels,in2010weconductedhumancrowdexperimentsattheTechnischeUniversitätBerlin.Oneparticularexperimentalsetup,forexample,arrangedfortwopedestrianflows(142and83subjects)tointersectatanangleof90degreesforoneminuteinaregionofabout25squaremeters,resultinginpeakdensitiesofaboutfourpedestrianspersquaremeter.

Thepedestrians‘spatio-temporalpositionswereobtainedviaphotogrammetricmeansfromvideodatarecordedwithmultipletemporallysynchronizednetworksurveillancevideocameras.TrackingofthepedestrianswasfacilitatedwiththestandardLucas-Kanadealgorithm.Inourcase,aparticularchallenge was presented by the fact that due to constructional limitations the scene could not be capturedfromabird‘seyeview.Thustheeffectofthepedestrians‘differentheightscouldnotbeassumedasnegligible-andsincethepedestrians‘heightswerenotknownbeforehand,weneededtodeviseamethodforextractingthepedestrians‘positionsreliablywithoutthisinformation.SmoothtrajectorieswerethenobtainedviaapproximationwithcubicB-splines,andthecombinatorialassignmentoftrajectoriesobtainedfromdifferentcameraperspectiveswassupportedwiththeKuhn-Munkresalgorithm.Wethencomputedalocaldensityfieldvianearest-neighborkerneldensityestimation.Comparedtothefixedbandwidthestimatorcommonlyusedintheliterature,ourapproachyieldsadensityfieldwhichprovidesspatiallyaveragedvaluesacrossmesoscopicregionsthataremorefaithful to the standard method of counting pedestrians in that region.

Wearguethatthespatio-temporaldensityconfigurationasarepresentationofpedestriandynamicsisparticularly suited for the calibration and validation of a variety of models: macroscopic simulations

alreadyproducedensityfields,anddataobtainedfromexperimentsormicroscopicsimulationsmaybeeasily converted.

SIMULATIONMODELSInthelastyears,severalmethodicalapproacheshavebeeninvestigatedforthemodelingandthesimulationoftrafficproblems.Inthepresentcontext,itseemsadequatetodevelopbothmicroscopicapproachesinwhichthepedestrianisconsideredasanindividualinteractingwithotherpedestrians,andmacroscopicmodelsinwhichpedestrianbehaviorisanalyzedintermsofmoreglobalpropertiesofacontinuousstream.Inthepresentprojecttwomicroscopicmodelsaredeveloped.Thefirstoneisagrid-basedapproachrootedincellularautomata(CA)models.Thesecondoneisacombinationofaforce based and graph based approach.

InthetraditionalCAmodelanditsvariousextensions,thestatechangeofthecell(i.e.position)isappliedtodescribethesystemdynamicsofthesimulation.Duetothisconceptuallimit,toourknowledge,thesimulationparticipants(i.e.pedestrians)areallassociatedwithafixedspacialsize,definedbythesizeofthegridcellintheCAmodel.Consequentlythepedestriansinthesimulationhaveafixedexclusivepersonalspacewhichdiffersfromempiricalobservations.Inourmodelthisexclusivepersonalspaceisgivenadditionalattention.Theeffectofamodifiableexclusivepersonalspaceisachievedbydefiningtheinaccessibilityofthesurroundingcellpositionofanarbitrarypedestrian.Bythismeansitispossibletodescribesimplegroupbehaviors,i.e.pedestrianswhichbelongtothesamegroupmayrequireasmallerexclusivepersonalspace,whiletowardotherpedestriansinthesimulationenvironment,thenearbycellpositionsaredeclaredasinaccessibleandthuskeepthelatteratarelativelylargerdistanceaswhatwewouldimagineinreal-worldsituations.Ourmodelalsopresentsanadvancedlocalstepcalculationtoenabletheso-calledmulti-cell-step,i.e. the transition from a start position to a destination with a distance larger than one grid cell. In thestepcalculationtheexecutionsequenceofthesimulationparticipantsisaffected,inadditiontotheparticipant‘sowncharacteristics,bytheactualsystemdynamicsaswell.Thisenablesasubstantialreductionofthe``deadlock‘‘phenomenon.Thismodelisappliedwithsomesimpleconfigurationswithwhichthepedestriansaregivenpre-definedstartpositionsanddestinations.Withthenotionofthemodifiablepersonalspace,thesimulationwithadvancedstepcalculationcanberealizedincombinationwithpedestriandensitycontrol,ifnecessary.

Thecombinedforce-andgraph-basedapproachtreatseveryparticipantasanagent,whomakesherowndecisions.Whilethegraphisonlyneededtofindapathfromtheorigintothedesireddestination,theagents‚Äôactuallocomotionisdrivenbyaforce-basemodel.Themodelnotonlyencompassessimplerepellingandattractingforcesbutalsomorecomplexforcesforexplicitcollisionavoidance.Themacroscopicapproachisbasedonasetofpedestrian-specificcoupledpartialdifferentialequations(PDEs)TheequationsarenotderivedfromtheEuler-/Navier-Stokes-PDEsknownfromfluidandgasdynamics.Thespecificsituationofmulti-destinationpedestriancrowdswithcrossingstreamsrequiresthedevelopmentofappropriatelyadaptedmethods.Thishasbeentargetedbytheuseofsimpleheuristics.

Typicalapplicationsoftheseapproachesincludereal-worldscenarioslikeairports,shoppingmalls,buildingsofmiddle-tolargesizeetc.,wheretheparticipants(i.e.thepedestrians)donotexhibitanoverallunanimityand(may)havedifferentandmultipledestinations.Beyondthemodelingoftheabove-mentionedproblems,aparticularaimofthisprojectwillbethedevelopment,implementationandtestofappropriatecomputer-basedsimulationmodels.Thereliabilityofthesemodelswillbeillustrated by a comparison with real data obtained from crossing pedestrian streams experiments.

Keynote


Mehdi Moussaïd, University Paul Sabatier, Toulouse FRANCE The modeling of human crowd behavior: Where physics meets cognitive science

Humancrowdsdisplayarichvarietyofself-organizedbehaviors,rangingfromthespontaneousorganizationoftrafficflowsundereverydaylifeconditions,totheemergenceofcrowdturbulenceatextremedensity.Tounderstandthemechanismsunderlyingthesephenomena,areliablemodelofpedestrianbehaviourisnecessary.Inthiscontribution,Iwillshow that such a model lies at the crossroad between physics and cognitive science.

First,Iwillshowthattwosimplecognitiverulesbasedonvisualinformationcandescribethemotionofpedestrianswell.Inparticular,thesetworulesaresufficienttoreproducetheself-organizedpropertiesofcrowdsobservedbelowacertaindensitythreshold.Asthedensityofpeopleincreases,however,bodycontactsbetweenneighboringpeopleoccur,andtheunderlyingmechanismschange.Duringovercrowding,intentionalavoidancebehaviorsarereplacedbyunintentionalphysicalinteractions,whichcanbedescribedbyusingNewtonianrepulsion forces.

Therefore,thelargevarietyofcrowdmovementsresultsfromadensity-dependantbalancebetweenphysicaleffectsandcognitiveprocesses.

Oral presentations


A.16 Ulrich Kemloh Wagoum, Forschungszentrum Jülich GmbH, Jülich GERMANY Empirical study and modelling of pedestrians’ route choice in a complex facility

A.17 Piotr Tofilo, Main School of Fire Service, Warsaw POLAND Staircase evacuation with firefighters counter flow - experimental data and output from popular modeling software

A.18 Robert Zinke, Friedrich-Schiller-Universität Jena, Jena GERMANY Psychological aspects of human dynamics in underground evacuation: field experiments

B.16 Gregor Lämmel, Technische Universität Berlin, Berlin GERMANY Getting out of the way: collision avoiding pedestrian models compared to the real world

B.17 Ignacio Martínez, Ineco S.A., Madrid SPAIN Methodology for pedestrian simulation with complex/random routes in public spaces

B.18 Ladji Adiaviakoye, Groupe ESEO, Angers FRANCE Collection of data stemming from the fine trajectory of the pedestrians

Empirical study and modelling of pedestrians’ route choice in a complex facility

Ulrich Kemloh Wagoum, Forschungszentrum Jülich GmbH, Jülich GERMANYArmin Seyfried, Bergische Universität Wuppertal,Wuppertal GERMANYFrank Fiedrich, Bergische Universität Wuppertal,Wuppertal GERMANYRalph Majer, Vitracom AG, Karlsruhe GERMANY

In recent years pedestrian dynamics has gained more importance and a lot of attention due to continuously growing urban population and cities combined with an increasing number of mass events.Majorissuesinthisareaincludeorientationandwayfinding.Thisisessentialforreproducingroutechoiceincomputermodelsandisdifficultduetothemanyunderlyingsubjectiveinfluencesonthischoice.Thispaperpresentsanempiricalanalysisofpedestrians’routechoicebehaviourinacomplexfacilitydonewithintheframeworkofareal-timeevacuationassistant[1].Thisstudyiscomplemented by a modelling approach used to reproduce the observed phenomena.

TheinvestigatedfacilityispartofthepromenadeoftheEspritArenainDüsseldorf,Germany.Theroute choice data are obtained from an automatic people counting system consisting of cameras. Thecamerasareoptimallydistributedatmainentrances,exits,andpassagesofthepromenade.Thisdistributionleadstoalogicalpartitioningoftheareainto5sections,whicharemappedtothedetectionareasoftheautomaticpeoplecountingsystem.Thecountingsystemconsistsof45monoand51stereocameras.Eachcameraismerelyresponsibleforoneexit,whichmeansthatateachtimetheinformation(inthiscasethenumber)aboutthepedestrianspassingthroughthatexitisavailable.Also,thepassingdirectionforeachexitisidentifiedmakingitpossibletoknowexactlyhowmanypedestriansareinsideaspecificsection.Thedataarepresentedintermsoffrequencies,i.e.thenumberofpersonspassingthecorrespondingcountinglineorexitperminuteinthedirection’in’and’out’ofasection.Usingthatinformationtheproportionalusageofthedifferentexitsofthepromenadecouldbecalculated.Thishasbeendonefordifferentfootballmatchesandconcerts.Theresultsfortwofootballmatchesplayedonthe5thand18thAugust2011arepresentedandanalyzedinthiscontribution.

Theframeworkusedfordescribingpedestriantrafficinthiscontributionisdividedinathree-tierstructure.Onedistinguishesbetweenthestrategic,thetacticalandtheoperationallevel[2].Atthestrategiclevelpedestrianschoosetheirdesiredfinaldestinationsandastrategytoreachthatdestination.Thestrategiesusedherearetheshortestpaths(localandglobal)combinedwithaquickestpathapproach.Short-termsdecisionsaretakenatthetacticallevel,avoidingjamsorswitchingtoafasterrouteforinstance.Basicrulesformotionsaredefinedattheoperationallevel.Theserulesincludeaccelerating,decelerating,andstopping.

Theoperationallevelisdefinedbythegeneralizedcentrifugalforcemodel[3].Thismodelbelongstothegroupofforcebasedmodelsandoperatesincontinuousspace.Pedestriansaredefinedbyellipseswithvelocitiesdependentsemi-axes.

Theunderlyingroutingmodelatthetacticalispresentedin[4].Itisaquickestpathapproachthatoperatesonanavigationgraph.Thegraphhasbeensemi-automaticallygeneratedfortheinvestigatedpartofthepromenade.Thishasbeendoneusingtheinter-visibilityofexits.Somenodeshavebeenmanually inserted into the initial graph when for instance two exits in the same room were not visible fromeachother.Afterthegraphisconstructed,wellestablishedalgorithmslikeFloyd–Warshallareusedtocomputetheshortestpathstoallpossiblefinaldestinations.Theinformationisstoredinordertosavecomputingresources.Thenodesofthegrapharenavigationlinesandthedesiredmovingdirectionofpedestriansisalwaysperpendiculartothoselines.Thevisibilityrangeofthepedestrians,whichdependsontheircurrentpositions,combinedwithanoptimalqueueselectionisusedfor(re)direction.Oncearrivedinanewlocation,thepedestrianssensetheirenvironmentandbasedontheirperceptions,takedecisionsthatminimizetheirtraveltimebysystematicallyavoidingjams.Thisisononehandachievedbyanalyzingtheprocessingspeedofdifferentqueuesatexitsandchoosingthebestsuitableone.Ontheotherhandpedestriansalsotrytoescapefromanalreadyexistingjamsituationwheneverpossible.Thisisthecasewhenapedestrianisnotconstrainedinthemiddleofacrowdforinstance.

Preliminaryresultsshowthatexitslocatedatcornersseemhiddenandthusarenotattractiveforpedestrians,buttheseresultsneedfurtherinvestigations.

REFERENCES

[1] S. Holl, A. Seyfried, Hermes - an Evacuation Assistant for Mass Events, inSiDe 7 (1) (2009) 60-61.URL http://inside.hlrs.de/pdfs/inSiDE_spring2009.pdf

[2] S. P. Hoogendoorn, P. Bovy, W. Daamen, Microscopic Pedestrian Wayfinding and Dynamics Modelling, in: M. Schreckenberg, S. Sharma (Eds.), Pedestrian and Evacuation Dynamics, 2002, pp. 123-155.

[3] M. Chraibi, U. Kemloh, A. Seyfried, A. Schadschneider, Force-based models of pedestrian dynamics, Networks and Heterogeneous Media 6 (3) (2011) 425-442. doi:10.3934/nhm.2011.6.425.

[4] A. U. Kemloh Wagoum, A. Seyfried, Optimizing the evacuation time of pedestrians in a graph-based navigation, in: M. Panda, U. Charraraj (Eds.), Developments in Road Transportation, Macmillian Publishers India Ltd, 2010, pp. 188-196

Staircase evacuation with firefighters counter flow - experimental data and output from popular modeling software

Piotr Tofilo, Main School of Fire Service, Warsaw POLANDMarcin Cisek, Main School of Fire Service, Warsaw POLANDKrzysztof Lacki, Main School of Fire Service, Warsaw POLAND

Thecounterflowconditionontheescapestairsisasituationthatmayoccurduringtheevacuationfrombuildingswhenthedownwardflowofevacuatingoccupantspassestheupwardflowoffirefightersheadingtothefirefloor.Itdoesnotalwaystakeplacebecauseinmajorityoflowerbuildingsevacuationstartsandendsbeforefirebrigadearrivesonthescene.Incaseoftallerbuildingstheconflictbetweenthetwomovementsismorelikelytohappenduetocertainfactors:thefirefightingandevacuationstrategyadoptedforthebuilding,theuseandpurposeofthebuilding,thenumberofstaircasesandsizingofthestair,firewarningsystemsavailable,thequalityofmanagementandstafftraining,theavailabilityoffirefightersliftsandthedistancetothefirestation.

Theevacuationstrategyintallerbuildingsisusuallyeithersimultaneousorphased.Incaseofsimultaneousevacuationthecounterflowconditionismorelikelyincaseoflongpre-movementtimesthatareusuallytheresultofinadequatewarningsystemsorpoortraining.Inthiscasethemovementsaremorelikelytocoincide.Animportantissuehereisthesizingofthestairsandthephilosophyherevariesslightlyfromcountrytocountry.ForexampleinUKthesizingofthestairsforsimultaneousevacuation is directly related to the whole number of occupants in the building which implies that the total number of people must have enough standing space within the staircase. In most countries howeverstairsaresizedtothehighestfloorpopulationbasedonassumptionthatfloorsarewellcompartmentedandonlythefirefloorandtheoneabovemustbeevacuatedimmediatelywhileotherfloorswillhavemoretime.Thisconceptisquiteclosetothephasedevacuationwhichallowsnarrowerstairs,howeverphasedevacuationisusuallygradualandcontrolledbywarningsystem,internalcommunication,welltrainedstaffandthefirecontrolcentre.Incaseofphasedevacuation,duetoadelayedevacuationofsomefloors,thelikelihoodandtheimpactofcounterflowincreases.

InBritishregulations(ADB)thisisacknowledgedbyarequirementforcheckingwhethertheeffectoffirefightersonphasedevacuationispossible.Thisshouldbetakenintoaccountwhiledecidingforthenumberofstaircasesinthebuildingassumingthatonemaybediscountedduetofirebrigadeactivities.Inpracticethemostlikelyoccasionswherecounterflowisexpectedaretalloffices,hotelsandthebuildingsthatarenottallenoughtohavefirefightingliftswhilehavingsignificantpopulation.Theproblem of counterflow can be even bigger in cases where disabled people are evacuated.

Theaimofthisworkistostudytheeffectsofcounterflowinconnectionwiththerequitementsofregulationsandmostcommonevacuationstrategies.Thesecondaimistoprovidedatafornumericalmodelingandcalculations.Toachievethesegoalsaseriesofexperimentswasconductedonastaircasewithandwithoutasimultaneouscounterflowofascendingfiremen.Experimentstookplaceinatypicalrepresentativestaircaseofanofficebuilding.Theeffectofthefirebrigadeinterventionon

evacuationmovementonstairswasexaminedforseveraldifferentinitialdensities.78peoplewereinvolvedtheexperiments–73asevacueesand5asfirefighters.Ineveryrunthesamenumberofstudentswasinvolvedbuttosimulatedifferentfloordensities,theywereplacedondifferentsegmentofthestaircase.Thelowerfloordensitywasappliedthemorelevelsofthestaircasewereoccupied.Thestartofeveryrunwassignalizedbyloudwhistleaudibleforallpeopleonthestaircase.Theevacueeswereaskedtomovecalmlyandnottorundownstairs.Theywerealsoaskedtomoveallatonceatthestartofeveryruntokeepthefloordensityfromthebeginningoftheruntothemomentwhentheymeetfirefighters.

Fordatacollectionafullmeasuringinstallationwasprepared.ItwascomposedoftwoIRdirectionalsensors (one for every exit door) connected with the counting unit to the computer with software. Thesensorsweremountedonthetopofbothdooropeningsandtheyweregivingimpulsetothecountingunitanytimewhenthepersonwaspassingthedoorintheoutsidedirection.Thesensorswerecalibratedtomeasureanybodywhoisatleast140cmtallandtheyshouldrecognizeeverypersoneveninhighfloordensity.Thecalculatingunitwassetuptogiveinformationtothecomputeraboutnumberofevacuatedpeopleinonesecondtimesteps.Additionallyvideocameraswereused.Onewasplacedonfirstfloorwhichwasthebestplacetoobservetheinfluenceoffirefightersmovementontheevacuationprocess.Thesecondvideocamerawasusedbyoneofthefirefighterstorecordthefireman‘sperspective during the counter flow movement.

Theresultsanddiscussionwillbepresentedinapaper.Afactoraccountingforthecounterflowisproposedforegressstairsizingcalculationsincaseswherecounterflowisidentifiedasapotentialproblem.Theexperimentalresultsarecomparedtosimulationsperformedwithcurrentpopularevacuationpackages:EVAC,STEPS,BuildingEXODUS,Pathfinder.Variousaspectsofsoftwareapplication,necessaryadjustmentsandtheimplicationsforevacuationstudiesinthecontextoffiresafety will be discussed.

Psychological aspects of human dynamics in underground evacuation: field experiments

Robert Zinke, Friedrich-Schiller-Universität Jena, Jena GERMANYGesine Hofinger, Friedrich-Schiller-Universität Jena, Jena GERMANYLaura Künzer, Friedrich-Schiller-Universität Jena, Jena GERMANY

Formodellingpedestrianbehaviour,especiallyinevacuationsituations,avarietyofsimulationsexists.Theyregardaspectsliketheenvironmentorthetypeofeventwithvaryingintensity;somealsoconsidersocialforcesandhumanfactorsaspects(overviewinSchadschneideretal.,2008).Ithasbeenclaimed for some time now that human factors aspects or social behaviour should be integrated in simulationmodels(e.g.Sime,1995).Anexampleisthesimulationofindividualagentswithdifferentemotionalstates(e.g.,Thiel-Clemen,2010).

Whilesimulationmodelshavestartedtoembraceabroadaccesstohumanfactors,empiricaldatafromlaboratoryexperimentsseemtoconcentrateonexternalaspectsofpedestrianbehaviour.Theseaspectsincludetherelevanceofthenumberofindividualmovingobjectsforthetimeneededforevacuation,andtheconsequencesofwalkwayordrivewaydimensions(e.g.Schreckenberg,2010;Schadschneideretal.,2009).Alsomoregeneralcharacteristicsofpedestrianmovementareinvestigatedwithevolvingpatternsandpathsofmovementsincrowdsunderlyingsocialforces(e.g.Moussaidetal.2011;Helbingetal.,2000).

Atthesametime,literaturecallsfortheinclusionoffurtherhumanfactorsaspectsintomodelsofevacuationbehaviour.Forexample,panicandirrationalbehaviourareinrealeventsnotfoundasoftenasexpected(so,Tubbs&Meacham,2007).Furtheraspectsaredistancekeepingbetweenpedestrians(Forell,2004)andaltruisticbehaviourinevacuationscenarios(Dynes,2006).Onlyfewfieldstudiesinvestigatethesephenomenaempirically;lessdosoinspecificinfrastructureslikeundergroundstations.Yet,onlyfieldstudiescandeterminetheincidenceandrelevanceofhumanfactors phenomena which the literature describes only in general. In order to identify observable humanfactorsandtodeterminetheireffects,fieldexperimentsarethemethodofchoice.TheresearchprojectOrGaMIRPLUS,fundedbytheGermanMinistryofEducationandResearch,deals with several aspects of human factors in evacuation of underground transportation systems. Byspecificallyaddressingsocialhumanfactorsaspectsofpedestrianbehaviourinthecontextofanundergroundsubwaystation,itmaycontributeforfurtherimprovingevacuationguidanceandsimulation models.

Theoverallaimistoimproveevacuationofstationsandtoshortenthetimeneededforsafeegressofpassengers.Basedonthefindingsfromfieldexperiments,observablehumanfactorswillbeformalizedaccordingtotheireffectsontheoveralltimeneededforevacuation.Apreliminarystudyinanundergroundsubwaystationincludedcountsofpassengers(e.g.howmanypassengersarehandicapped,travelincompanyorgroups).Structuredinterviews(n=213)withpassengersaskedforknowledgeaboutevacuation,warningsignals,andinterviewees’estimation

oftheirownbehaviourinanundergroundevacuation.Oneresultisaclearpreference(67%)ofnotleaving without one’s partner or children. Interviewees also said they wouldn’t use an unfamiliar way out if instructed to do so.

Basedontheresultsfromthepreliminarystudyandtheaspectsfoundintheliterature,fourseriesoffieldexperimentswerecarriedoutindifferentundergroundtrainstationsinGermany.Thefieldexperimentsvariedseveralinfrastructural,individualandsocialvariables.Groupsizesvariedbetween30and130.Tasksincludedmovingquicklyalongdifferentdistancesontheplatform,upanddownturnedoffescalatorsandstairs,andegressingontracksintothetunnel.Structuredinterviewswereconductedaftereachsingleexperiment.Alltaskswerevideo-recordedandlateranalysedforre-occurringsocialandhumanfactorsaspects.Participantswereaskedtocommentontheirownbehaviour,themovementofthegroup,andhowtheyexperiencedinfrastructuralspecificities.Firstresultsshowthatinfrastructuralelements,groupsizeandindividualvariations(handicaps,preliminaryexperiencewithsimilarsituationsorinfrastructure)allinfluencedthespeedofegress,speedingituporslowingitdown.Especially,aninfluenceof(social)humanfactorswasfound.Examplesaresupportingthemovementofhandicappedorslowerpersons,stayingwithone’sgrouporrelyingonguidanceofothersfamiliarwiththeinfrastructuralspecificities.Also,spontaneous“knots”,asaresultofindividualfactorsorsocialadhesionofgroups,werefound.

Theseresultsarerelevantforundergroundevacuationbecauseevacuationofpassengersisnotonlysloweddownbyarchitecturalbottlenecksbutalsobyinfrastructuralfactors,individualandsocialfactors.Infurtherfieldexperimentsphenomenawillbeinvestigatedwithlargernumbersofpassengers,forpossibleculturalvariationanddifferenttypesofinfrastructures.

REFERENCES

Berlonghi, A. (1993). Understanding and planning for different spectator crowds. In Smith, R. and Dickie, J. (Eds.). (1993). Engineering for crowd safety (S.13-20). Amsterdam: Elsevier.

Clarke, L. (2002). Panic: Myth or reality? Contexts, 1 (3), 21-26.

Dynes, R.R. (2006). Panic and the vision of collective incompetence. Natural hazard observer, Vol. XXXI (2).

Fahy, R. F. & Proulx, G. (2009). ‚Panic‘ and human behaviour in fire. Paper presented at the 4th International Symposium on Human Behaviour in Fire, Robinson College, Cambridge.

Forell, B. (2004). Bewertung der Anforderungen der Musterversammlungsstättenverordnung (Mai 2002) hinsichtlich realistischer Evakuierungsszenarien in Diskotheken und ähnlichen Vergnügungsstätten. Braunschweig: TU Braunschweig.

Helbing, D., Farkas, I., and Vicsek, T., Freezing by heating in a driven mesoscopic system, Physical Review Letters 84, 1240-1243 (2000).

M. Moussaïd, M., Helbing, D. and Theraulaz, G. (2011) How simple rules determine pedestrian behavior and crowd disasters. PNAS 108 (17) 6884-6888.

Schadschneider, A., Klingsch, W., Klüpfel, H., Kretz, T., Rogsch, C. and Seyfried, A. (2009). Evacuation dynamics: Empirical results, modeling and applications. In B. Meyers (Hrsg.), Encyclopedia of Complexity and System Science (S. 517-550). Berlin: Springer.

Schreckenberg, M. (2010). Dynamik von Menschenmassen, AKNZ Seminar, 22.06.2010. Deutsche Hochschule der Polizei, Brühl.

Sime, J. D. (1995). Crowd psychology and engineering. Safety Science, 21(1), 1-14.

Thiel-Clemen, T., Klingenberg, A. (2010). Kombination von zielorientiertem Verhalten und Emotionen in Individuen-orientierten Simulationen, in: Wittmann, J. and Maretis, D. (Hrsg.), Simulation in den Umwelt- und Geowissenschaften. Osnabrück: Shaker. 71-80.

Tubbs, J. S. and Meacham, B. J. (2007). Egress design solutions: A guide to evacuation and crowd management planning. Hoboken, NJ: John Wiley and Sons.

Getting out of the way: collision avoiding pedestrian models compared to the real world

Gregor Lämmel, Technische Universität Berlin, Berlin GERMANYMatthias Plaue, Technische Universität Berlin, Berlin GERMANY

Numericalsimulationofhumancrowdsisachallengingtask,andanumberofapproachestosimulatepedestriandynamicsonamicroscopiclevelhavebeenestablished.Theaimistomodelarealistic,andinparticularcollision-free,movementofcrowdsincomplexenvironments.

OnepossibleapproachtoachievethisgoalareCellularAutomata(CA).CAmodelsrepresenttheenvironmentwithagrid-likestructure,whereeachcellofthegridmaycontainatmostonepedestrianatatime.CAmodelshaveoftenbeenusedforthesimulationofevacuationscenarios.AcommonproblemwithexistingCAmodels,however,isthattheydonotmodelcomplexwayfinding,sinceinmostCAmodelswayfindingisimplementedviaagloballydefinedpotentialfield.Intheory,itwouldbepossibletoassignanindividualpotentialfieldtoeachpedestrian.Thisapproach,however,wouldbetocomplex in terms of computational costs for large scenarios.

Othersimulationconceptsuse(discretized)differentialequationssimilartoequationsknownfromthedescriptionofMolecularDynamics(MD).ProbablythebestknownmodelbasedontheMDanalogyisthesocialforcemodel.Insocialforcemodelsimulations,eachpedestrianhasadesiredvelocitytowardsadesireddestinationandadaptshercurrentvelocityaccordingly.Apedestrian‘sincentivetoavoidobstaclessuchasotherpedestriansismodeledbyrepellingforces.Force-basedmodelsarewellunderstood and have reasonable computational costs.

Thereisathirdclassofmodelsthattrytoachievecollisionfreepedestrianmovementincomplexenvironments.Thesemodelsarebasedonthesocalledconfigurationspaceobstacleapproachandhavetheirfoundationinrobotics.Inthiscontexttheconfigurationspacedescribesallpossiblelocationsapedestriancanreach.Locationsthatcannotbereachedarethesocalledconfigurationspace obstacles. For this approach the pedestrians and the obstacles in the environment (e.g. walls and otherpedestrians)havetoberepresentedasasetofsimplepolygons.ApaththroughtheenvironmentiscollisionfreeifthepathdoesnotintersecttheMinkowskisumofthepolygonalobstacleswiththepolygonal representation of the pedestrian reflected in her reference point.

Anextensiontotheconfigurationspaceobstacleapproachisthevelocityobstacleapproach.Similartotheconfigurationspaceobstaclesthevelocityobstaclesdescribeallvelocitiesapedestriancanchoosethat will lead to a collision at some point in time assuming straight movement and no acceleration of rest of the pedestrians. In the velocity obstacle approach every pedestrian chooses at each point intimeavelocitythatavoidscollisionandisclosetothedesiredvelocity.Thisworkinvestigatestwoapproachesforexplicitcollisionavoidanceformulti-destinationpedestriancrowdssimulationsincomplex dynamic environments.

Thefirstapproachisanextensiontoanexistingcollisionavoidingforce-basedmodel(Zanlungoetal.,2011).Intheforce-basedmodeltherepellingforcesdonotonlydependonthelocationsoftheobstaclesandpedestriansbutalsoontheirvelocities.Whencalculatingtheactualforcethataffectsagivenpedestrianthemodelfirstcalculatestheminimaltimeofclosestapproachtoallotherpedestriansandobstaclesintheenvironment.Allpedestriansarethanprojecteduptothispointintimeassumingconstantmovement.Theinfluencingforceonthepedestrianinquestioniscalculatedbased on this projection and weighted by the inverse of the minimal time of closest approaches times. Thisleadstoabehaviorthatalwaystriestoavoidthenextpotentialcollisionundertheassumptionofconstantmovement.Likeinotherforce-basedmodels,too,eachpedestrianhasadesiredvelocitytowhichsheadaptsinagiventimeiffreemovementispossible.Inourapproach,complexwayfindingis implemented via a navigation graph and a shortest path search: a force that lets a pedestrian move alongthenavigationgraphreplacesthesimpledesiredvelocityvector.Withthismodelitispossibletosimulate pedestrian movements to multiple destinations in a complex environment.

Thesecondapproachdiscussedinthispaperisbasedonthevelocityobstacleapproach(seev.d.Bergetall.,2008).Butinsteadofcalculatingvalidvelocitiesduringeverytimestepthepedestrianschoosevalidaccelerationspreferablyclosetothedesiredone.Therefore,werefertoithereasaccelerationobstaclemodel.Thisapproachleadstothesameresultaschoosingavalidvelocitydirectlybutletsthemodelbebetterintegratedwithforce-basedmodels.Thisintegrationisneededsincethecalculationofthedesiredvelocityandtheshort-rangeinteractionisstillbasedonforces.Likewithourfirstapproach,desired movement is based on a force that lets each agent move along a navigation graph.

Bothmodelsaretestedondatafromareal-worldexperimentconductedbyus(see,Plaueetal.,2011).Theparticipantsoftheexperimenthavebeendividedintotwogroups,andeachofthegroupswasinstructedtowalkalongagivenpath.Thesepathswerearrangedsuchthatthetwopedestriangroupsintersectatanangleofabout90degrees.Theexperimenthasbeenrecordedandtheindividualtrajectorieshavebeenextractedafterwardsfromvideo.Inthispapertheperformanceoftheforce-basedmodelandtheaccelerationobstaclemodelareevaluatedbycomparisonwiththereal-worlddataintermsofthespatial-temporaldistributionofpedestriandensities.Tothisend,wecalculatealocaldensityfieldviakerneldensityestimationsimilarto(Helbingetal.,2007)butwithvariablekernelbandwidth.Afurtherinvestigationdiscussedinthispaperistheperformanceofbothapproachesintermsofcomputationalcosts.Attheendofthepaperafinalappraisalisgivenwhichofthemodelsismost appropriate under which conditions.

REFERENCES

J. van den Berg, M. Lin and D. Manocha: „Reciprocal Velocity Obstacles for Real-Time Multi-Agent Navigation“, IEEE International Conference on Robotics and Automation (ICRA 08): IEEE, pp. 1928-1935

D. Helbing, A. Johansson, and H. Z. Al-Abideen: „Dynamics of crowd disasters: An empirical study“, Phys. Rev. E 75, pp. 046109 (2007)

M. Plaue, M. Chen, G. Bärwolff, and H. Schwandt: „Trajectory extraction and density analysis of intersecting pedestrian flows from video recordings“, Proc. PIA11, LNCS 6952, 285-296 (2011)

F. Zanlungo, T. Ikeda, and T, Kanda: „Social force model with explicit collision prediction“, EPL (Europhysics Letters), 2011, 93, 68005

Methodology for pedestrian simulation with complex/random routes in public spaces

Ignacio Martínez, Ineco S.A., Madrid SPAIN

Usually,duringthefunctionaldesignorevaluationofapublicspace,itisessentialtohaveatoolthatguaranteesthesuitableoperationofthedesignedspaces.Modernrailwaystations,forinstance,areincreasingly becoming sophisticated spaces that combine the commercial plurality of a shopping center with the complex boarding processes of an airport.

Inthiskindofspaces-airports,stations,intermodaltransportationhubsetc.-peoplefindahighvarietyofwaypointsorintermediateactivitiesintheirjourney(suchasrestaurants,shops,baths,cashdispensers,standsorscreens,amongothers).Thosewaypointsusuallyreachthehundred,orevenmore.Insuchcases,thenumberofpossibleroutesforpedestriansalongtheirjourneytowardstheirfinaldestinationgrowsexponentially.Thisleadstoahugecombinationoffeasiblepathways.Modellingpedestrianmethodologiessofarused,areusuallybasedonaO/D-basedfixedpercentageofroutes,thatallocatespeopleintheminordertomovearoundthemodeledspace,throughoneorseveralintermediateactivitiesbeforetheyreachtheirfinaldestination.Theseconventionalandsimplemethodsare,therefore,nolongerenoughtorepresentthecomplexityofsuchspaces.Theyarejustunabletoreproducethecomplexityofreal-timedecisionmakingforpedestrians.

Thus,itisimportanttodevelopandimplementamethodologyabletoreproduceandassessthosecomplexpedestrianmovementsinsidethiskindofspaces.Thismethodologyhastobesimpleenoughasnottobetoocumbersomeorcalculation-intensiveforacommonPC,ifwewantittobepractical.Beingsaidthis,itisnecessarytoleavebehindconventionalsolutionsbasedonsimplisticapproachestotherealsituation,andtofindamethodpreciseenoughtobecomeapracticaltoolinthedecision-makingprocess.

Onthebasisoftheseprinciples,ithasbeendevelopedamethodologywhichreproducesthebehaviorofpeopleinpublicspaces.Fromadeepknowledgeofcurrentsituation,peoplebehaviorhasbeenanalyzedandprocessedineachcase,effectivelyidentifyingbehavior-patternsthatareusedasthebasisforvalidationofthesepedestrianmodels,andforthedesignoffuturescenariosaswell.ThismethodhasalreadybeensuccessfullyappliedinthetwomajorrailwaystationsinMadrid,ChamartínStationandAtochaStationfortheSpanishRailwayInfrastructureAdministrator(Adif).

Thekeycomponentofthismethodologyinvolvesimportantfieldsurveyscampaign:O/Dmatrix,countsandcommutertracking,usedtoidentifyallthesteps/waypointsofusersinsidethesevenues.Thosesurveyshaveachievedmorethan5,000commuterstrackingsinChamartínStationandmorethan11,000trackingstocommutersinAtochaStation.Thosetrackingsfollowpeoplefromthemomenttheyenterthestudiedspacetothemomenttheyleft,coveringalloperatinghoursandintermediate

stops. Final reports were expanded to the total daily demand so every user in the station had a register linewithexhaustiveinformationdataabouttheirwaypoint“chain”.Thetrackingsincludedothernecessaryinformationtopreciselycharacterizeeveryusersuchas:• Timetheuserenterthescenario.• No.ofentrance.• Movement1.• TimetheuserisarrivingmovementNo.1.• TimetheuserisleavingmovementNo.1.• Movement“i”.• TimetheuserisarrivingmovementNo.“i”.• TimetheuserisleavingmovementNo.“i”.• Finaldestination.• Trainschedule.• Trucknumber.• Timetheuserreachesthefinaldestination.• No.ofexit.• Luggage(shoulderbag,bag,smallsuitcase,largesuitcaseetc.)• Agerange.• Sexoftheuser.• Typeofuser(boardingtravelers,passingusers,transitusersetc.)

PassengerdailydemandofChamartínStationin2010wasover39,000people,andAtochaStationexceeded215,000boardingpassengersaday.Havinginmindthosefigures,itisclearthataspecificmethodologyisrequiredinordertoeffectivelysolvetheproblemofanalyzingtheirbehavior,anditssubsequentapplicationinapedestrianmodel.

InthecaseofChamartínandAtochaStations,pedestrianmodelswerebuiltusingLegionsoftware.Theflexibilityofthissoftwarepackageallowedtheimplementationoftheroutingmethodology,redirectingpeopleinsidethestationthroughtheirprobabilisticrouteswithoutknowingtheminadvance.

Sincetheroutesdependonthetypeofpassenger,timeoftheday,andtypeofwaypoints(restorationareas,ticketmachines,waitingareas,etc.),itprovidednotonlyagreattooltocalibrate/validatearealisticmodel,buttoassessthebehaviorofthemodelinfuturescenariosvaryingthespaceallocationfor the commercial and operational areas.

Asawhole,thedevelopmentofaspecificmethodologyallowedtosolvetheproblemoftheanalysisofimportantdatainthetrackingcampaigndatabasefromthefieldsurveyanditsdirectapplicationinthedevelopment,calibrationandvalidationofthepedestrianmodel,akeyelementinthedecision-makingprocess.

Collection of data stemming from the fine trajectory of the pedestrians

Ladji Adiaviakoye, Groupe ESEO, Angers FRANCEPatrick Plainchault, Groupe ESEO, Angers FRANCE

Oneofthemainofascientificapproachistounderstandwhatisreal.Forthis,wemustidentifythephenomenawhichcouldthatexplainindifferentways.Onewaytostudythesephenomenaistobuildamodel,inordertosimplifytherepresentationoftheobservedphenomenatoberelevant.Tobuilarelevantmodel,wechoosesomerepresentativedataandwemadeanumberofassumptionsaboutthebehaviorofthephenomenastudied.Aspartofastudyonthebehaviorofpedestrians,mosttheoreticalmodels(e.g.tohelpunderstandanorganizationorsocio-spatialdynamics)arefarfromrealistic.Businessmodels(e.g.fortownandcountryplanningdecisionmaking)orpredictivemodels(e.g.totrytopredictfuturedevelopmentofcities)lackofdataexperimentaltobevalidatedexperimentally.Thisworkisapartofaprojecttoexperimentallyvalidatesomepedestrianmodelwithexperimentaldata.Inordertoknowthestrategyadoptedbypedestrianswhentheygetaround.Weneedtoanalyzeand to understand the physical interactions(movement and force between individuals) among a crowd. Themechanismsthatunderliethedynamicsofhumancrowdsarelittleknown.Theyareratherstudiedinpsychologyandsociology,toanalyzetherelationshipbetweenpedestrians(SilcockD.,R.Walker,1998).Thisanalysisismainlythesubjectofaqualitativestudy.

Anotherfocusforpedestriansistoconsidermeasurementsinrealconditions.Forinstance,inahall,itcouldbeinterestingtomeasurethepathtakenbypedestrians,butalsothespeedortheacceleration.Infactitwouldallowadetailedanalysisoftheapproachofthestride.Aconsequentlyananalysisratherquantitative.Ourworkplaceinthiscontextandnarrowlycombineobservationsinnaturalenvironmentwithcontrolledexperimentations.Currentlywedevelopahighprecisionwhosepurposeistotrackthemovinginanindoorarea.Thelatterwouldbewidelysufficienttorepresentinterestingspaceforresearchersinpedestrianmobility.Theassociationofspeedandpositionsensors,likeinputdevicesofthisreal-timesystem,shouldimprovetrackingapplications.Asaresultseveralapplicationscouldbecarriedout:crowdtracking,trackingofconsumersinshoppingmalls,trackinginevacuationcases,topredictthequalityofpedestriantrafficindifferentconditions,toevaluatetherightarchitecture for capacity of buildings.

Ourprojectreliesonliteraturebasedonobservationsanddescriptionsofpedestrianbehavior(K.Teknomo,2002)(SPHoogendoorn,2004).Twopedestrianapproachessetupthetrackingsystem.-Kinematic:Resultsofstudiesinthisareaarenothomogeneous(TFNovacheck,1998)(T.Fugger,Jr.,B.Randles,2000)(A.Toor,A.Happer,2001),butthesenumericvaluesgiveagoodindicationoftheordersofmagnitude.Hencetheynevermustbeconsideredlikeabsolutereferences.-Mechanical:Thepedestrianisacomplexstructureof41degreesoffreedom(A.Ebel,D.Hanon,2002.).Hecanbemodeledbyamechanicalsystemwithsixdegreesoffreedom(3inrotationand3intranslation).

Bothkinematics(changesaccordingtothedensityofthecrowd)andmechanicalstructure(mustbetakenintoaccountlikearigidobjecttoreducethecalculatingpowerofoursystem)ofthepedestrianwillhavetobeconsideredtoanswertosystemspecificationsofourproject.Tocombinethesetwoapproaches,weselectedsomerepresentativedataandwemadeanumberofassumptionsonthedensitiesofcrowds.Fromtheseassumptions,wedeterminedadensityof1.1personpersquaremetererrorofabout5cmontheposition.Thesystemmusthavearefreshrateof4Hzforagoodtrajectory.

Intermsofinteractionwiththeenvironment,coordinationandorganizationofcollectivemovementsresult from interactions between individuals but also the interactions that individuals establish with theirenvironment.Theenvironmentwillinfluencethemovementofindividualsactingeitherontheindividualhimself,oronthesignal(tactile,acoustic,visual).Itisimportanttonotethatindividualswillalsoactontheirenvironmentinreturn.Wefocusedinitiallyontheroleoftheenvironmentonthesemechanismsbydevelopingahighlycollaborativeandnon-intrusive.Pedestriansarenotawarefollowed,oraslittleaspossibletobe,soiftheymoveintheusualway.

Thefactofbeingbasedonknowledgeofphysics,meansthatwewillnotmakeanyassumptionaboutthe mode of human intelligence to be the source of orders that will cause the movement of pedestrians. Wesimplychooseasystembasedonthedataavailableonthesysteminputonly.Theuseofpressurecarpet or infrared stereo vision seem better suited to the problem.

Inthedetectionstep,theabilityofthecarpettofeelthepressureperpixelallowsthefloortolocateandanalyzethespeedanddirectionchangesinthepedestrianenvironment.Identificationandtrackingofusersbystereocameraallowanalysisofinteractionswithhighprecision.Oneofthetwosystemswillberetainedandpresentedafteralaboratorytest.Inaddition,somesocialfactorsthatmaycauseembarrassment or that can inspire social interaction around the pedestrian will be explored.

Thispaperthereforepresentsatrackingsystemcapableofstoringdetailedkinematicsofpedestrianbehavior in an environment. Its mechanisms are consistent with studies in the physical sciences on the behaviorofpedestrians.Oursystemfostersanindividualizedstudy,eachpedestrianistreatedasanindividualandtreatedasaseparateentity.Itis,moreover,abletomemorizethekinematicsofseveralhundred people in real time.

REFERENCES

[1] A. Ebel, D. Hanon, B. Stanciulescu, P. Pudlo, E. Grislin, F-X. Lepoutre (2002). Modèled‘animation comportemental de piétons virtuels

[2] A. Toor, A. Happer, R. Overgaard, and R. Johal, “Real world walking speeds of youngpedestrians,” in SAE, 2001.

[3] D. Silcock, R. Walker, and T. Selby, “Pedestrians at risk,” in PTRC proceedings of European transport conference, 1998, pp. 209–219.

[4] K. Teknomo, “Microscopic pedestrian flow characteristics : Development of an image processing data collection and simulation model,” Ph.D. dissertation, Departement of Human Social Information Sciences, Graduate School of Information Sciences, Tokohu University, Japan, Mar. 2002.

[5] S. P. Hoogendoorn, “Pedestrian modelling by optimal control and differential games,” in German - Dutch - Finnish Seminar on Traffic Engineering, Feb. 2004.

[6] T. F. Novacheck, “The biomechanics of running,” Gait & Posture, vol. 7, pp. 77–95, 1998.

[7] T. Fugger, Jr., B. Randles, A. Stein, W. Whiting, and B. Gallagher, “Analysisof pedestrian gait and perception-reaction at signal-controlled crosswalk intersections,” Transportation Research Record 1705, pp. 20–25, 2000. [Online]. Available :http ://www.enhancements.org/trb%5C1705-004.pdf

Keynote


Chris Kemp, Buckinghamshire New University, Uxbridge UNITED KINGDOM New developments in crowd management, safety and dynamics: Managing risk for safer crowded places

InmytalkIwillpresentaseriesofaspectsfocusingoncrowdmanagement,safetyanddynamicsthatfocusonthemanagingriskforcrowdedplaces.Inrecentyearsthedevelopmentof modelling for crowded spaces has become more realistic in its development and through a range of activities such modelling can become even more honed so that the delivery of realtimeandrealisticmovementsacrossarangeofenvironmentscanbeenhanced.Thefirstpart of the presentation will focus on the underpinning traits of play associated with the crowdedspace.Thesecondpartlooksattheusesofprofilingtools.Thethirdpartreviewsthemanagement of the space itself.

Thefinalpartsofthepresentationfocusonvoronoimodellingandthecreatingofactivitieslinkedtocrowdedspacesandfinishesbylookingattheeverpresentthreatswhichneedtobemodeled to understand the total environment encompassed within a crowded space.

Special session


Dinesh Manocha, University of North Carolina, Chapel Hill NC USA Computer vision, graphics and robotics

Fromrecord-settingcrowdsatralliesandproteststofuturisticswarmsofrobots,ourworldiscurrentlyexperiencingacontinuingriseofcomplex,distributedcollectionsofindependentlyactingentities.Withpotentialapplicationssuchascomputergraphics,predictingcrowddisasters,improvingrobotcooperation,andenablingthenextgenerationofairtravel,developingmodelstoreproduce,control,predictandunderstandthesetypes of systems is becoming critically important.

Inthistalk,Iwillgiveanoverviewofhowtousevelocity-spaceplanningtechniquestocompute cooperative motion paths for a group of independent entities sharing the same physicalspace.Iwillfocusonthespecialcaseofsimulatinghuman-likecrowds,withapplicationstocomputeranimationandarchitecturalanalysis.Specifictopicswillincludeoptimization-basedstrategiesfordistributedcollisionavoidance,usesoftheprincipleofleasteffortforsimulatingcrowds,anddata-drivenstrategiesformodelingdifferencesinpersonalities.Thetalkwillalsocoverrelatedtechniquesneededtoachieveaccuratesimulationsoflarge-scalecrowdssuchasefficientparallel/SIMDcomputemodelsandmethodsofvalidatingsimulationsagainstrealworlddataandwilldiscusshowvelocity-spacemotionplanning can be applied to collision avoidance for distributed robotic systems.

Mubarak Shah, University of Central Florida, Orlando FL USA Computer vision, graphics and robotics

InthistalkIwillpresentourrecentcomputervisionmethodforhumandetectionandtrackinginavideosequence.Ourapproachlearnspart-basedperson-specificSVMclassifierswhich capture the articulations of the human bodies in dynamically changing appearance and background.

Withthepart-basedmodel,ourapproachisabletohandlepartialocclusionsinboththedetectionandthetrackingstages.Inthedetectionstage,weselectthesubsetofpartswhichmaximizestheprobabilityofdetection,whichsignificantlyimprovesthedetectionperformanceincrowdedscenes.Inthetrackingstage,wedynamicallyhandleocclusionsbydistributingthescoreofthelearnedpersonclassifieramongitscorrespondingparts,whichallowsustodetectandpredictpartialocclusions,andpreventtheperformanceoftheclassifiersfrombeingdegraded.

Keynote

Friday, 8 June - 09.00

Andrew Hutton, Network Rail, London UNITED KINGDOM London Bridge – the significant role ped modelling plays in designing the new station and how to deliver it

ThispresentationwilltakedelegatesthroughthevariousstepsweusedpedestrianmodellingforindetermininganacceptabledesignforthenewLondonBridgestation.Itwillhighlighthowweconsideredtheexistingdemand,understoodtheoperationalconstraintsandfactoredthoseagainstthedesignaspirationsandrequirements,whichincludedagrowthprofileofover60%.Thiswasthensubjectedtobothstaticanddynamicmodellingtovalidatethedesigns.

Thisworkhasbeenverywellreceivedbyboththeclientandstakeholders,whogainedgreatconfidenceintheprocess.Therefore,averysimilarapproachisalsobeingusedtoconsideranyproposed alterations to the design as well as all impacts that the eight large construction phases willbringtothestation.AlltheworkhastobecarriedoutwhilstmaintainingtherailwayoperationsofthefourthlargeststationintheUK,whichhandles1000peopleperminuteinthepeakhour!

Oral presentations

Friday, 8 June - 09.30

A.19 Hubert Klüpfel, TraffGo HT GmbH, Flensburg GERMANY PedGo Guardian: an assistant for evacuation decision making

A.20 Qiuping Li, Wuhan University, Wuhan CHINA Ant colony based evacuation optimization algorithm for mixed vehicle-pedestrian flows

A.21 Kerry L. Marsh, University of Connecticut, Storrs CT USA Crowd guidance in building emergencies: Using virtual reality experiments to confirm macroscopic mathematical modeling of psychological variables

B.19 Anders Johansson, University College London, London UNITED KINGDOM Utilizing crowd insights to refine disease-spreading models

B.20 Sean Curtis, University of North Carolina, Chapel Hill NC USA Pedestrian simulation using geometric reasoning in velocity space

B.21 Daichi Yanagisawa, College of Science, Ibaraki University, Ibaraki JAPAN Influence of rhythm and velocity variance on pedestrian flow

C.19 Tao Chen, Tsinghua University, Beijing CHINA Study of human behaviour before evacuation

C.20 Elise Miller-Hooks, University of Maryland, College Park MD USA Modeling pedestrian route choice during large public gatherings

C.21 Alexander Mordvintsev, National Research University of Information Technologies, Mechanics and Optics, St. Petersberg RUSSIA Simulation of city evacuation coupled to flood dynamics

PedGo Guardian: an assistant for evacuation decision making

Tim Meyer-König, TraffGo HT GmbH, Flensburg GERMANYHubert Klüpfel, TraffGo HT GmbH, Flensburg GERMANY

Inmodernsocieties,urbanizationandlargescaleentertainmentareobvioustrendsandconsequently,largeandcomplexmulti-functionalarenasarepartofmostlargecities.Theguidanceofpedestrianflowsinsuchbuildingsisachallenge.TheHermesresearchproject,fundedbytheGermanFederalMinistryofEducationandResearchwithintheSecurityResearchProgram,aimstoprotectandsavelives by developing an “evacuation assistant” that could allow stadiums and their venues hosting large eventstobeclearedquicklyandmoresafelythancurrentlypossible.TheHermessystemisdesignedtouse information about a current situation to predict what will be the future positions of the occupants. Itinvolvesfeedingdataintoacomputerprogramabouttheavailabilityofrescueroutes,fireprotectionsystems,andthedistributionofpeopleasdeterminedusingvideocameraslinkedtoimage-processingsoftware.Thisallowsforflexiblereactionstotheactualsituationforwhichusuallynospecialemergency plans exist.

Inthiscontribution,wewillfocusonthe“Guardian”modewhichhasbeenportedtothePedGoevacuation simulation software.

DESIGNOFEVACUATIONSYSTEMSThedimensionsofescaperoutesarespecifiedinbuildingdesigncodes.SincethereferencecaseistheESPRitarenainDüsseldorf,Germany,wemakereferencetotheGermanregulations(“Versammlungsstättenverordnung”,resp.SBauVO,part1).Foranon-roofedsportsstadium,awidthof1.20mper600persons(minimum:1.20m,increasinginstepsof0.60m)isrequired.Forroofedsportsstadiumstherequirementsare:1.20mper200persons,andanadditional0.60mforevery100additionalpersons).Inadditiontothat,differentmaximumdistancesarespecifiedsuchastoa“safearea”(whichmightbeafireprotectedstaircase).

Thegeneraldesignprocessforpedestriantrafficalsoappliestorouteelementsofanevacuationsystem.Asmentionedabove,thedesigncriteriaaccordingtotheprescriptivecodesarethelengthandwidthofescaperouteelements.Therefore,prescriptivecodesarestaticapproachesnottakingintoaccountthecrowddynamicsaspectsofanemergencyevacuation.TheHermesevacuationassistantandconsequentlyalsoPedGoGuardianontheotherhand,arenotonlydesigntoolsbutalsocovertheoperationofthestadium,e.g.thecrowdmanagement.Tothisend,videoanalysistechniqueslikedetection and tracing are used to adjust the simulation and provide information about the current positions of all persons in the arena.

Basedontheforecastofthesimulation,thesecuritystaffcanprepareforeventstocomeandcantestdifferentstrategiestohandleapotentiallyhazardousorveryuncomfortablesituation.Inordertoassess

thedifferentstrategiesandtheirconsequences,alevelofhazardconceptbasedonthelevelofserviceand the crowd density and movement characteristics is used.

ThelevelsAtoFaregroupedintothreebroadercategories.“Green”,“yellow”,and“red”arethelevelsusedintheevacuationcontext.FortheHermesproject,thismorebasicschemeisadoptedtomakeitapplicable for a command and control center.

EVACUATIONSIMULATIONEvacuationsimulationisacorepartofanevacuationassistant.FastmicroscopicsimulationslikePedGocanpredictthemotionofmorethan50,0000personseasilyinrealtimeonstandardDesktoporLaptopcomputers.Theydotakeintoaccountindividualpropertiesandpersonaldifferencesconceningphysiological(likewalkingspeed)andpsychologicalcharacteristics(likeorientation).Theinformationof the detection system about the position of all persons in the arena is processed by the simulation kernel.Similartoaweatherforecast,thekernelthencomputesaprognosisforthesituationinthefuture.Forthecaseofanevacuation,thisforecastcoversthecompleteprocess,i.e.itrunsuntilthelastperson has reached a safe place of refuge.

Anexamplefortheinitialsituationforafootballmatchwithallpersonsplacedinitiallyonthestandsisshowninfigure(willbepartofthefullpaper).Thereddotsindicatepersonssittingontheirseats.Thecumulativelocaldensityisonemajorcriterionforidentifyingcongestion.Infigurextheareaswherethiscumulativelocaldensityexceeds3.5personspersquaremeterformorethan10%oftheoverallegresstimeismarkedred.

EVACUATIONASSISTANTTheevacuationassistantwillbeusedbythestaffinthecommandandcontrolcentertoassessthesituationandplantheoperationofthesecuritystaffandforces.

Thedensitycriterionshowninthetableaboveisonlyoneforassessingcriticalsituations.Thegreen/yellow/redschemewillassistthesecuritypersonneltofocustheirattention.Aslongasasituationis“green”,nospecificattentionisnecessary.Yellowrequiresspecialattentionandpreparationforcrowdmanagementmeasuresandotherinterventions.Finally,redsituationsarenotacceptableandrequirespecificandimmediateactionlikecrowdcontrol.Ofcourse,thechallengeisthedefinitionofthedemarcationcriteriabetweenthedifferentregimesandthecalibrationoftheinterventionandactioncatalogue.

Ant colony based evacuation optimization algorithm for mixed vehicle-pedestrian flows

Qiuping Li, Wuhan University, Wuhan CHINAZhixiang Fang, Wuhan University, Wuhan CHINAQingquan Li, Wuhan University, Wuhan CHINA

RESEARCHMOTIVATIONAlthoughmulti-agentsimulationtoolsareincreasinglyusedfortransportationsimulationinrecentyears,effectivemechanismstosimulatethemixedvehicle-pedestrianflowsundercongestedevacuationsituationsstillremainsatitsinfancy.Differentfromvehicles,pedestriandoesnotfollowstricttrafficrulesandtheyspontaneouslystop,changedirectionsormakesuddenturns.Duetorealworldsituations,especiallyinthedevelopingcountries(i.e.ChinaandIndian),weshouldresearchthemixedvehicle–pedestrianflowsinevacuation.Thetraditionaltransportationsimulationmodelsusuallyfocusonvehiclesintheroadnetworksorthepedestriansinthebuildingenvironment.Viswalk,anewlydevelopedsimulationmodule,canbeintegratedwithVissimtosimulatetheinteractionsbetweenpedestriansandvehicles(Viswalk,2011).However,approachestointegratethesimulationmodelwiththeoptimizationmodelforplanningtheoptimalmulti-modalevacuationprocessarestillveryfew.Inthispaper,weproposeatwo-tierarchitecturetosimulatethemixedvehicle-pedestrianflowandoptimizetheevacuationprocess.Onetierisdesignedtofindpathofvehiclesandpedestriansoptimizedbyantcolonyalgorithm(ACA),andtheothertierisforsimulatingthemovementsofvehicles and pedestrians dynamically.

METHODThefirsttierisfortheoptimizationpurpose.Itcalculatesthepathswhichpedestriansorvehicleswillmovealongstepbystep.EachpedestrianorvehicleisviewedasanantinACA.Thetotallengthofspace-timepathsforallpedestriansandvehiclesisconsideredastheoptimalobjective.Thisobjectivecaninvolveboththeinfluenceofevacuationtimeandevacuationpathlength.Weuseaprobabilityselectionandapositivefeedbackstrategytofindtheoptimalpathsinemergencysituation.Ineachstep,allthepossiblepathsforeachvehicleandpedestrianarefoundandassignedtothemaccordingtoaroulettewheelselectionmethod.Thetrafficflowcalculatedinthelaststepwillguidethenext-steppathfindingofantorantgroupinthemicro-simulationprocess.Afteracertaintimesofiteration,iftheobjectivevalueremainsrelativelystable,thepathsfoundbytheantsforpedestriansandvehiclescanbeviewedasthesystemoptimalpathsinthemulti-modalevacuationsituations.

Thesecondtiersimulatesthedynamicmovingprocessofvehiclesandpedestrians.Mixedflowsinlinksandintersectionsofanevacuationnetworkaremodeledandsimulatedseparately.Inthelinks,thisstudysimulatesthepedestrianflowandvehicleflowwiththeseparatedqueuemodel.Forexample,vehiclesmovesonvehiclelanesandpedestrianswalksonfootpaths.Lateralfrictionisignoredbetweenthesetwotrafficmodes.Attheintersections,thespaceisdividedbycells.Apedestrianoccupiesonecellandavehicletakesseveralcells,buteachcellcanonlybeusedbyavehicleorapedestrian.Thebasicruleforananttochooseacellisbasedontheleasttraveltimetothiscell.Theconflictofdifferent

pedestriansorvehiclescompetingforonecellissolvedbyFCFS(firstcome,firstserve)principle.Baseonthesimulationresults(e.g.linkdensityandconflictdegreeofpedestriansandvehicles,etc.),eachvehicleorpedestriancanfindareasonablepathintheevacuationnetworkaccordingtothetime-dependenttrafficflow.

EXPERIMENTRESULTAnexperimentonastadiumevacuationisimplemented.Thepeopleinthestadiumisdividedintotwoclasses,oneclassneedtoescapethestadiumbywalking,andtheotherclasswanttopickuptheircarsanddrivetoescapefromthestadiumarea.Theareawithin1kmradiusfromthethestadiumisconsideredastheevacuationnetwork.Theevacuationnetworkinthisstudyincludes272nodesand644linkswithdirections.Theproposedtwo-tierarchitecturehasbeenimplementedinC++languageandrunsonaPCwith3.06GHZ,3GBRAM.TheACArunsfor100generations,andtheaveragespace-timelengthgoesdowngraduallyfrom4020inthe1stiterationtoabout3000in40thiteration,andthentheevolutionprocesskeepsrelativelystable.

CONCLUSIONInthispaper,weproposedatwo-tierarchitecturetointegratethesimulationmodelwiththeoptimizationmodelforplanningtheoptimalmulti-modalevacuationprocess.Becauseofthesignificantdifferenceinmovingrules,traditionaltransportationsimulationmodelshavesomedifficultiesinmodelingthedynamicsofbothpedestriansandvehicles,especiallyinmodelingtheinteractionsofthesetwotravelmodes.Moreover,thenetworkMOEs(measureofeffectives)couldbeoverestimatedifweignoretheinteractionsbetweendifferenttravelmodes.Ourexperimentresultsshowthatourapproachisusefulintheevacuationplanoptimizationforpedestrian-vehiclemixedflowsinevacuationnetworks.However,itmustbenotedthattheproposedmicro-simulationmodelfor dynamics of pedestrians and vehicles needs to be further improved since the results obtained by the proposedmodelcannotbevalidatedduetoalackofexperimentalinformationonhumanperformanceunder emergency condition.

REFERENCE

Viswalk(2011): http://www.vissim.de/software/transportation-planning-traffic-engineering/software-system-solutions/viswalk/ (accessed on 2011-11-10)

Crowd guidance in building emergencies: Using virtual reality experiments to confirm macroscopic mathematical modeling of psychological variables

Kerry L. Marsh, University of Connecticut, Storrs CT USAChristian Wilkie, University of Connecticut, Storrs CT USAPeter B. Luh, University of Connecticut, Storrs CT USAZhenxiang Zhang, University of Connecticut, Storrs CT USATimothy Gifford, University of Connecticut, Storrs CT USANeal Olderman, University of Connecticut, Storrs CT USA

Crowdevacuationbehaviorsincludingdisorderandblockinghavebeenobservedinrecenttragediessuchasthe2003RhodeIslandand2009Bangkoknightclubfires[1][2].Behavioralstudiesofevacueeshaveshownthatpsychologicalstressplaysanimportantpartintheemergenceofdisorderandblocking[3][4].However,thereisadiscrepancybetweentheoriesthatexplainthebehaviorofevacueesandthemethodsofprovidingeffectiveguidancetoevacueesinbuildingemergencies.Withrecentadvancesinfiredetectionmethodsandcrowdcommunication,thereispotentialtoalleviatethesekindsofinjuriesanddeathsinthefuture.Thusitisimportanttoredressthisgapbetweentheoryandourabilitytoguidecrowdstosafety.Acriticalissueistheelucidatingimportantpsychologicalfactorsthatinfluenceegressforprovidingeffectiveguidance.

Toaddresstheproblemofprovidingeffectiveguidancetocrowds,anoptimizationproblemwasformulatedinourpreviouswork[5].Theunderlyingequationswerechosentofollowamacroscopicmodel,wherecrowdsaretreatedasafluid[6][7]toallowfastoptimization.Ourmodelimprovesupon these models by integrating psychological phenomena that previously have been examined only withincomputation-intensivemicroscopicmodels.Inparticular,onenovelparameter,thedesiredflowrate(evacuees’feelingofurgencytomove),wasdevelopedasamacroscopiccounterpointtoHelbing’sdesiredvelocity[8].Thiscanhelpexplaintheemergenceofdisorderandblockingduringanemergencyevent.Theeventualgoalistosolvethisprobleminreal-timetherebyprovidingeffectiveguidancetoevacueesduringanactualemergencyevent.Althoughthecurrentstateofcrowdguidancefallsfarshortofsuchanobjective,validationofourmodelisamajorsteptowardsrealizingthisgoal.

Inthispaperweanalyzetheeffectsofthepsychologicalfactorsthatareinourmodel,aswellasexplorethepotentialvalueofincludingotherfactors.Topartiallyvalidateourexistingmodel,anexperimentwasconductedtoconfirmwhatthemodelimpliesaboutthepsychologicalexperienceofevacueesinsuchsituation.Todothis,avirtualrealitytestingplatformwascreated.Thisallowedustorunmanytrialswithparticipantsandtobetterrecreatethepsychologicaleffectsofanemergencyeventbymakingtheevacuationanembodied,immersiveexperience.Forthevirtualrealityenvironment,auniversitylibrarywaschosenduetothepotentialbenefitsofguidanceatthislocation.Inparticular,thislibraryisoftencrowdedwithcollegestudentswhoareunfamiliarwithemergencyexits.A3DmodelofonefloorwasconstructedandusedintheevacuationsimulatorFireDynamicsSimulatorwithEvacuation(FDS+Evac)[9]tofindtheevacuationroutesandspeedsofagents.GraphicalavatarswerecreatedwiththeirroutesandspeedsimportedfromFDS+Evacanddisplayedduringtheexperiment.Theparticipantthusexperiencedtheevacuationfromafirst-personperspective.

Tobetterrecreatetheperceptionofbeinginarealemergency,participantsworeahead-mounteddisplay and were moved along the same route and speed as a random simulated agent (based on the FDS+Evacoutput).SeveralFDS+Evacevacuationsimulationswererunwithvariedunimpededwalkingspeeds.Havingawiderangeofpossiblespeedsallowsustodeterminethedesiredspeedforaparticipantmovedatseveraldifferentspeedsondifferenttrials.Thethreeunimpededwalkingspeedswereuniformlydistributedinmeters/secondasslowU(0.325,0.925),mediumU(0.95,1.55),andfastU(1.3,3.7),notingthattheunimpededwalkingspeedforadultsisU(0.95,1.55)inFDS+Evac[9].Inthisvirtualenvironmenthazardousconditionswerealsovariedbydisplayingdifferentcombinationsofsmokeandfire.Additionally,duringeachevacuationtrial,fireandemergencyalarmsoundswereheardthroughaspeaker.Thus,throughthismanipulationoftheenvironmentweattemptedtorecreateapsychologicallyrealisticemergencyeventfortheparticipant.Participantsgavereal-timecontinuousfeedbackontheirrelativedesiredspeedduringeachtrialbymovingajoystickforwardorbackward.Joystickpositionwasexpressedasavaluebetween1and7,withvaluesbelow4indicatingdesiretomoveslowerandvaluesabove4indicatingdesiretomovefaster.

Analysesindicatedthathazardconditionhadastatisticallysignificanteffectonjoystickposition.Inthetrialsinvolvingnovisiblehazards,i.e.,nofireorsmoke,themeanpositionwas4.801.Whensmokewasvisible,themeanpositionwas5.204.Whenbothsmokeandfirewerevisible,themeanvaluewas5.717.Thusdesiredspeedincreasedwhenhazardswerevisible,indicatingtheeffectivenessofthevirtualenvironmenttoinduceappropriatepsychologicalresponses.Additionalfindingswhichindicatetheeffectivenessofthevirtualenvironmentarethechangesindesiredspeedduetothechangesinwalkingspeed.

Giventhepromisingresultsofthesefindings,weplantotestotherpsychologicalfactorsincludingexcessiveurgencytomoveduetoblocking,trustinsocialversusnonsocialinformation,andtheeffectsofleadershipandsocialbonds.Wearealsoworkingonperformingfiredrillsinthesamelibraryourvirtualenvironmentiscreatedfrom.FurtherresultswillbepresentedattheConferenceandincludedinthefinalversionofthepaper.

REFERENCES

[1] W. L. Grosshandler, N. P. Bryner, D. Madrzykowski and K. Kuntz, “Report of the technical investigation of the Station Nightclub fire,” NIST NCSTAR 2, Vol. 1, 2005.

[2] S. Mydans, „At least 59 die in Bangkok club fire,“ New York Times, Retrieved on 2009-01-01.

[3] G. Proulx, “A stress model for people facing a fire,” Journal of Environmental Psychology, Vol. 13, No. 2, 1993, pp. 137-147.

[4] R. F. Fahy and G. Proulx, “Human behavior in the World Trade Center evacuation,” Fire Safety Science - Proceedings of the Fifth International Symposium, Melbourne, Australia, 1997, pp. 713-724.

[5] P. Wang, P. B. Luh, S. C. Chang, J. Sun, “Modeling and optimization of crowd guidance for building emergency evacuation,” Proceedings of the 2008 IEEE Conference on Automation Science and Engineering, Washington DC, August 2008, pp. 328-334. DOI: 10.1109/COASE.2008.4626553.

[6] L. G. Chalmet; R. L. Francis; P. B. Saunders, Network Models for Building Evacuation, Management Science, Vol. 28, No. 1, Jan., 1982, pp. 86-105.

[7] H. W. Hamacher and S. A. Tjandra, Mathematical Modeling of Evacuation Problems – A State of the Art, in Pedestrian and Evacuation Dynamics, M. Schreckenberg and S. D. Sharma, Ed., Springer, Berlin, 2002, pp. 227–266.

[8] D. Helbing, I. Farkas and T. Vicsek, “Simulating dynamical features of escape panic,” Nature, Vol. 407, 2000, pp. 487– 490.

[9] T. Korhonen and S. Hostikka, “Technical Reference and User’s Guide for Fire Dynamics Simulator with Evacuation, FDS+Evac,” VTT Technical Research Centre of Finland, April 2009, http:// www.vtt.fi/ proj/fdsevac/fdsevac_downloads.jsp.

Utilizing crowd insights to refine disease-spreading models

Anders Johansson, University College London, London UNITED KINGDOM

Increasing urbanisation has been accompanied by the rise of ever larger world cities with higher population densities as it becomes more obvious that the economies of scale generated by urban agglomerationleadtoincreasedprosperity.Bigcitieshavebecomethegeneratorsofwealthastheyincreasinglyattracttheworld’smoreskilledpopulationsthroughgreatstreamsofmigration.

Apartfromthestrainthatsuchgrowthisputtingonurbanlivingandtheprovisionofurbaninfrastructure,astheplanetgetsincreasinglycrowded,citiesinparticulararebecomingplacesofmorefrequentandlargermassgatherings,resultinginmassivecongestiononroadsystemsandpublictransport,whilelocalentertainmenteventsgenerateextremecrowdinginsmallspacessuchassportsarenas,festivalsandotherpopularentertainmentsites.Thisextremecrowdingisparticularlyproblematic during emergency evacuations.

Therearemanypositivesocialandeconomicbenefitsofbringingpeopletogether,butitiswidelyacknowledgedthattherearealsoseveralnegativeoutcomes.Whenthedensityofpeople(asdefinedasthenumberofpersonspersquare-meter)ofpeoplegrowstoohigh,increasedcrime,severetrafficdelays,andpollutionaregenerated,oftenmorethanproportionatelythroughtheinteractionofpopulations.

Denselypopulatedareasarealsoidealmediaforcertainrespiratoryepidemicstodevelopandspread,duetotheproximityofpeople.Frequentinteractionsbetweenpeoplewhosephysicalcontactincreasesnon-linearlywiththenumberlocatedinanyparticularplace,are,asiswellknown,locationswherecontagious diseases spread rapidly and reach the largest population in the shortest time.

Eventhoughepidemiologicalprocessesarecloselyrelatedtopedestriancrowdingandothermodesoftransport,thetimescalesaretypicallylongerandthespatialextentsarelargeraswell.Forthesereasons,epidemiologicalmodelstypicallyoperateonamacroscopicsystemlevelratherthanonamicroscopicindividual-basedlevel.Theadvantageofworkingatamacroscopiclevelisthatthescaleoftheproblemdoesnotbecomealimitingfactor.Buttherearealsodisadvantages,fortheinterventionsthatonecanmaketopreventorlimitdiseasefromspreading,typicallyoperateonamicroscopicindividual-basedlevel.Examplesoftheseareimmunization,screening,puttingpeopleintoquarantine,and even setting up some travel restrictions which have to be individually tailored to the travellers in question.

Moreover macroscopic models are typically based on the assumption that populations are in equilibrium,arehomogenous,andwell-mixed,whichweknowisnottrueforreal-worldpopulations.

Mobilityandinteractionpatternsinrealpopulations,asreflectedinthewaythatcitiesareorganized,tendtobehighlyskewedformsofdistributionmorelikepower-lawsratherthanbeinguniformornormallyGaussiandistributed.Thistypicallymeansthatdistributionsofthepopulationcontainclusters on all scales but with few really large clusters and a large number of very small clusters. Diseaseswillspreadfasterinthelargestclustersbutwilltendtodieoutwhentheclustersizeisreachedandthishasimportanteffectsontheactualpatterningofthespreadofanyepidemic.

Forallthesereasons,weproposeashiftinepidemiologicalmodellingfromthemoretop-downmacroscopiclevel[1]tothemicroscopicbottom-up[2,3].Therehaverecentlybeensomeattemptstomoveepidemiologicalmodelstoamicroscopiclevel,forexamplebyrunningcomputersimulationsofthespreadofcomputervirusesinscale-freenetworkssuchastheInternet[4],orinusingtheinternationalairlinetransportationnetworktogetherwithcensuspopulationdatatosimulatethespread[5].WeproposetogoevendeeperintothemicroscopiclevelandmakeuseofavailabletrajectoriesofindividualsobtainedviatechniquessuchasGPSormobilephonetracking[6],andinthisway,begintotrackhowindividualsinsmallspacesandinclosedtransportsystemssuchassubwaytrains enable the spread of disease due to their proximity.

Wedemonstrateexamplesofsuchsimulations,showinghowanindividual-basedcrowdmodelcanmirrormoreaggregateSusceptible-Infected-Recovered(SIR)modelsthathavedominatedthefieldsofar[1].

REFERENCES

[1] Kermack WO, McKendrick AG. A Contribution to the Mathematical Theory of Epidemics. Proc. Roy. Soc. Lond. A. 1927; 115, 700—721

[2] Johansson, A., Batty, M., Hayashi, K., Albar, O., Memish, Z., and Marcozzi, D. (2011) Crowd and environmental management during mass gatherings, The Lancet Infectious Diseases, in press

[3] Khan, K., McNabb, S., Memish, Z., Eckhardt, R., Hu, W., Kossowsky, D., Sears, J., Arino, J., Johansson, A., Barbeschi, M., McCloskey, B., Henry, B., Cetron, M., and Brownstein, J. S. (2011) Mass gatherings: Infectious disease surveillance and modeling across geographic frontiers and scientific disciplines, The Lancet Infectious Diseases, in press

[4] Pastor-Satorras R, Vespignani A. Epidemic Spreading in Scale-Free Networks. Phys. Rev. Lett. 2001; 86: 3200–3203

[5] Colizza V, Barrat A, Barthélemy M, Vespignani A. The role of the airline transportation network in the prediction and predictability of global epidemics. PNAS 2006; 103 (7): 2015-2020

[6] González MC, Hidalgo CA, Barabási A—L. Understanding individual human mobility patterns. Nature 2008; 453, 779—782

Pedestrian simulation using geometric reasoning in velocity space

Sean Curtis, University of North Carolina, Chapel Hill NC USAJur van den Berg, University of Utah, Salt Lake City UT USAStephen J. Guy, University of North Carolina, Chapel Hill NC USAJamie Snape, University of North Carolina, Chapel Hill NC USAMing Lin, University of North Carolina, Chapel Hill NC USADinesh Manocha, University of North Carolina, Chapel Hill NC USA

Pedestriandynamicshaslargelybeendominatedbycellularautomata-andsocialforce-basedmethods.Theseapproacheseachhavetheirownuniquepropertiesandcosts;tovaryingdegreestheyreproduceobservablepedestrianphenomenabutbearcostsinherentintheirparticularformulation.Wepresentanovel,robustmicroscopicmodelforsimulatingpedestriansandtheirinteractionswhichcapturesthesameaggregatephenomenaexhibitedbypreviousapproaches.Furthermore,thisnewmodelencompassesalargerspaceofpedestrianbehaviorandcanbeimplementedinanumericallyefficientand stable manner.

Cellularautomata(CA)simulatespedestriansbydiscretizingtheworkspaceintocells.Acellisoccupiedbyatmostonepedestrian.Pedestriansmovefromcelltocellbasedonasetofprobabilitiesappliedtoneighboringcellsandthenupdatingagentssequentiallyandapplyingcollision-resolutionruleswhenmultipleagentsseektomoveintothesamecell.CAhasbeenshowntoexhibitmanyemergentphenomena(e.g.laneformation,jamming,etc.)

Thesocialforce(SF)modelsimulatescrowdsofpedestriansasaparticlesimulation.Pedestrianshavemassandareattractedtowardagoalbyadrivingforce.Repulsiveforcespreventpedestriansfrompassingthrougheachotherandobstacles.Theparticle’svelocityandpositionresultfromtheintegrationoftheaccelerationimpartedbytheforces.LikeCA,SFgivesrisetoemergentphenomena(laneformation,jamming,etc.).SFformulationscanalsoexhibitadherencetothefundamentaldiagram[1].UnlikeCA,SFoperatesinacontinuousspaceofpositionandvelocity;thesimulatedpedestrians can be arbitrarily heterogeneous. Wepresentanovelagent-basedmodelofpedestrianmovement:reciprocalvelocityobstacles.Weassert that humans have the ability to estimate speed and future positions of other pedestrians. Considertwopedestrians,AandB.WhenAobservesB,AestimatesthepathofB.IfAdeterminesacollisionisprobable,Awilladjustitspath,otherwise,Awillcontinuealongitsoriginalpath.Reciprocalvelocityobstacles(RVO)modelthisestimationandadaptation.

RVOhasitsorigininrobotics.Avelocityobstacle(VO)isageometricconstructioninvelocityspace.PedestrianBcreatesavelocityobstacleforA,whichcorrespondstothesetofallvelocitiesthatAcouldtakewhichwouldleadtoacollisionwithB(assumingconstantvelocityforB)[4].EachneighborofAlikewisedefinesanadditionalvelocityobstacleforA.GivenA‘spreferredvelocity(thevelocitytheagentwouldtakeintheabsenceofothers),A‘scollision-freevelocityisthevelocitylyingoutsidetheunionofvelocityobstacleswhich„best“approximatesthedesiredvelocity.

Usingvelocityobstaclesdirectly,eachagentwouldviewotheragentsasunresponsivedynamicobstacles.However,theotheragentsare,infact,responsive.Thisoversimplificationcaneasilyleadtooscillationsbecauseeachagentoverreacts.Furthermore,itisnotconsistentwithobservedhumanbehaviorinwhichbothadapttheirpathstoavoidcollisions[2].TheRECIPROCALvelocityobstaclecapturesthisnaturalbehavior.Inourmodel,thechangeinrelativevelocityrequiredtoavoidcollisioniscomputedandapportionedbetweentheagentsequallysobothagentsexertefforttoavoidcollision.Furthermore,thisreciprocitycanbemodeledinsuchawayastoguaranteemutuallycompatible,collision-freevelocitiesbetweentheagents[4].

Ithasbeenobservedthatasthedensityofagroupofpedestriansincreases,theaveragespeedofthatgroupdecreases.Thisrelationshipiscalledthefundamentaldiagram.Thisphenomenonhasseveralorigins.First,biomechanicalresearchhasshownthatwalkingstridelengthgrowswithspeed.Densecrowdsadmitlessspaceforsteppingwhichleadstolowerspeeds.Second,pedestrianestimationoftheavailablespacewillinvariablycontainerror.Therefore,thepedestrianstendtoguessconservatively.Finally,thenatureofthatconservativeguessisdependentonaculturalcomponent.Someculturesimplicitlyvaluespacemorethanothers.Wehaveencodedthesethreefactorsintoourmodelandthe resulting simulation adheres to the fundamental diagram over a broad spectrum of physical and culturalvariation.Finally,aswiththepreviousapproaches,RVO-basedpedestriansimulationexhibitsawiderangeofemergentphenomenasuchaslaneformation,arching,edgeeffects,etc. RVO-basedpedestrianscontrastwellwithCAandSF.IthasbeenobservedthatCAgeneratesmacro-levelbehaviorsconsistentwithrealcrowdsofpedestrians,butatthemicrolevel(thelevelofindividualpedestrians)thetrajectoriesare“notrealistic“[3].TrajectoriesproducedbyRVOaresmoothandphysicallyreasonable.Furthermore,CA‘sabstractionhomogenizesthepedestrians;allagentsmoveatthesame,fixedspeed.RVO,likeSF,allowsforaheterogeneouspopulation. RVOalsocomparesfavorablywithSF.TherepulsiveforcesusedbySFtopreventcollisionsleadtoapotentiallystiffsystem.Asthecrowdbecomesdense,smallertimesteps(ontheorderof0.01-0.001s)arenecessarytoproducestablesimulationresults.RVOperformscomputationsforcollision-freevelocitiesinvelocityspace.Assuch,RVO-basedpedestriansimulationisstableforquitelargetimesteps(empiricalevidencehasshownsuccesswithtimestepsaslargeas0.2s.)AuthorsofSF-basedsystems have observed that simulation parameters must be specially tuned for each novel scenario [1].ForRVO,experiencehasshownthatasinglesetoffixedparametervaluesspansalargerangeofscenarios.Finally,whileSFcangeneratereasonablepedestrianbehaviors,therearereasonable,observablepedestrianbehaviorswhichSFcannotgenerate.Forexample,picturetwopedestrians,AandB,travelingonperpendicularpaths,headedtowardaninevitablecollision.OnerealisticstrategyBcouldapplytoavoidthecollisionistoincreaseitsspeedinitsdesireddirectiontopassinfrontofA.SFcannotreproducethisbecausetherepulsiveforcefromAactingonBwillalwayshaveacomponentthatiscountertoB‘spreferreddirectionandcanonlyservetoretardB’sspeedinthatdirection.RVOcomputesa„best“alternativeofthepreferredvelocity.Thisalternativecouldlieinalargespaceofresponsesincluding,butnotlimitedto:speedinguptocrossinfront,slowingtoallowtheotherpedestriantopass,turning,etc. Reciprocalvelocityobstaclesserveasanappealingmicroscopicmodelforagent-basedpedestriansimulation.TheRVOformulationencodesthehumanabilitytopredictandadapt.GroupsofpedestrianssimulatedwithRVOexhibitemergentbehaviorsandconformtothefundamentaldiagram.Inpractice,theimplementationisefficientandrobust.Collision-freevelocitiescanbecomputedforthousandsofagentsinafewmillisecondsonasinglecoreofaCPU.

REFERENCES

[1] M. Chraibi, A. Seyfried, and A. Schadschneider. Generalized centrifugal-force model for pedestrian dynamics. Phys. Rev. E, 82(4):046111, 2010.

[2] J. Pettré, J. Ondre, A. Oliver, A. Cretual, and S. Donikian. Experiment-based modeling, simulation and validation of interactions between virtual walkers. In Proc. Symposium on Computer Animation (2009), ACM.

[3] S. Sarmady, F. Haron, and A. Talib. A cellular automata model for circular movements of pedestrians during tawaf. Simulation Modelling Practice and Theory, 2010

[4] J. van den Berg, S. J. Guy, M. Lin, and D. Manocha. Reciprocal n-body collision avoidance. In Inter. Symp. on Robotics Research, 2009.

Influence of rhythm and velocity variance on pedestrian flow

Daichi Yanagisawa, College of Science, Ibaraki University, Ibaraki JAPANAkiyasu Tomoeda, Meiji Univeristy / JST, CREST, Tokyo JAPANKatsuhiro Nishinari, University of Tokyo, Tokyo JAPAN

INTRODUCTIONANDABSTRACTImportantgoalsofpedestrian-dynamicsresearchalongwithelucidationofcollectivephenomenaaredevelopmentofsolutionstoeasecongestionandcontributiontothesafety.Thus,wewilldevelopamethod to increase pedestrian flow in the congested situation in this paper.

Weobtainaflow-densitydiagram(FDD),whichdepictstherelationbetweendensityandflowbyconsideringtheeffectoflengthofstrideandpaceofwalking.Itindicatesthatwhenpedestrianswalkinaconstantpaceirrespectivetothedensity,theflowincreasesinthehigh-densityregionevenifthepaceisslowerthanthatinthenormal-walkingcondition.Wehaveperformedtherealexperimentandverifiedthatthepedestrianflowincreaseswhentheirwalkingpaceiscontrolledbyslowrhythm.

Theexperimentalresultalsoindicatesthattherhythmdecreasethevarianceofthepedestrians’movement.Thus,weextendourmodeltodealthemeanandthevarianceofthevelocityofpedestriansindependently.Inthecasethatthemeanvelocityofeachpedestrianissame,flowincreaseswhenthevariancebecomessmallinaone-dimensionalcircuit.Bycontrast,flowdecreaseswhenthevariancebecomes small in a simple evacuation model.

STRIDEANDPACEFUNCTIONWeconsiderthevelocityVofanindividualpedestrianindetailbydividingitintotwopartsasV=SP,whereS(stridefunction)andP(pacefunction)denotelengthofastrideandpaceofwalking(totalnumberofrightandleftstepsperunittime),respectively.

Theexplicitformulationofthestridefunctionisintuitivelydeterminedasfollows.Itisplausibletoassumethatthereisthemaximumlengthofstrideforpedestriansinthelow-densityregime.Whenthedensitybecomeslarge,pedestriansarenomoreabletowalkwiththeirmaximumstrideanditdecreases linearly as the headway distance decreases.

Theexplicitformulationofthepacefunctionisdeterminedasfollows.Ifthedensityislowandapedestriandoesnotinteractwitheachother,itisfeasibletoassumethatpedestrianswalkwithconstantpace.However,contrarytothestridefunction,itisdifficulttoobtaintheexplicitformulationofthepacefunctioninthehigh-densityregimewithsomeintuitiveassumptions,thus,weconsiderasimplelinearfunctionandinvestigatehowthechangeofpaceaffectsontheflow.

PEDESTRIANFLOWThepedestrianflowiscalculatedasQ=rV=rSP,whererisdensity.Thesecondderivativeoftheflow

indicatesthatconvexityoftheflowQinthehighdensityregimeisdominatedbythecharacteristicofthepacefunction,i.e.,whetherthepaceincreases,decreases,orremainsconstantaccordingtotheincrease of the density.

IMPROVEMENTOFFLOWBYSLOWRHYTHMHerewewouldliketonewlyproposeasolutiontoimproveflowinthecongestedsituationfromourmodel,inwhichthevelocityofpedestriansisrepresentedbytheproductofstrideSandpaceP.Ifpedestrianswalkwithaconstantrhythm,inotherwordsifwecancontrolthewalkingpacebyrhythmusingadevicesuchasametronome,therhythmexactlycorrespondstothepace.Therefore,fastandslowrhythmsincreaseanddecreasetheflow,respectively.

Weassumethatthepaceofpedestriansdecreasesinthehighdensityregime,i.e.,theFDDisconvexdownwardinthenormal-walkingcase.Thenwesurprisinglyobservethattheflowoftheslow-rhythmicwalkingcasebecomelargerthanthenormalwalkingcase.Thisphenomenonmaygiveasafety solution to ease congestion in the real world since the flow increases without any excessive haste.

EXPERIMENTWehaveperformedtheexperiment,wherepedestrianswalkintheone-dimensionalcircuit.Twokindsofwalkingwereperformedintheexperiment.Inthefirstcase,wedidnotgiveanyspecificinstructionstotheparticipants,sothattheywalkednormally.Inthelattercase,theparticipantswereinstructedtowalkwiththesoundfromtheelectricmetronome,whoserhythmis70beatsperminute(BPM).Notethatwedidnotuniformwhichfoottomovefirst.Theexperimentalresultsindicatethefollowings:1.Weseethattheflowislargerinthenormalcasethantherhythmiccaseinthelow-densityregime,sothatwehaveverifiedthat70BPMisslowerthanthenormal-walingpace.2.Theflowincreaseslinearlyasthedensityincreasesinthelow-densityregimeinthebothcases,sothatwehaveverifiedthatparticipantswalkedwithconstantstrideandpace.3.Theflowdecreaseslinearlyinthehigh-densityregimeintherhythmiccaseasexpectedfromouranalysis.4.Weobservethattheflowinthenormalcaseisconvexdownwardinthehigh-densityregimeaswehave assumed in the theoretical analysis.5.Sincethetheoreticalassumptionsoftheconvexityaresatisfiedintheexperimentalflows,thecrossingappears,andtheflowoftherhythmiccaseexceedsthatofthenormalcaseinthehigh-densityregime.Therefore,wehavesucceededtoverifythatslowrhythmimprovesthepedestrianflow.

EFFECTOFVARIANCETheexperimentalerrorsaresmall;however,weobservethattheyaresmallerintherhythmic-walkingcasethaninthenormal-walkingcaseinthehigh-densityregime.Thisresultimpliesthatrhythmremovestheheterogeneityofpedestrians‘movement,synchronizesit,andcontributestothehomogeneousspatialdistribution,whichachievesimprovementoftheflow.

Inordertoverifythisphenomenon,wehaveextendedourmodeltodealthemeanandthevarianceofpedestrians’velocityindependentlybyapplyingaunitconversionofphysicalquantitiesinstochasticcellular-automatonmodelswithdiscretetime.Wecanfocusontheinfluenceofvarianceontheflowbyremovingtheinfluenceofmeanusingthismethod.Ourtheoreticalresultsindicatethatflowincreaseswhenthevarianceofvelocitydecreasesinthecasethatpedestrianswalkinaone-dimensionalcircuit.Bycontrast,whenpedestriansevacuatethroughanarrowexit,decreaseofthevelocityvariancedecreases the flow since plural pedestrians try to go through the exit at the same time and conflict there.

Study of human behaviour before evacuation

Tao Chen, Tsinghua University, Beijing CHINALili Pan, Tsinghua University, Beijing CHINAGuoquan Zhang, IKEA China, Beijing CHINA

Computersimulationisthemainmethodtostudyevacuationduringfireofotherdisasters.Intheevacuationsimulation,oneofthemostimportantinitialparametersistheoccupantcharacteristics,suchastheage,gender,education,distribution,accompaniedstatusandpre-movementtime.Inpractice,theseparametersarebasedonassumptionandexperience.However,duetolackofevacuationandpedestriandata,validationorcalibrationofevacuationmodelsisstillachallengingproblemwhichhasnotgotagoodsolution.ThisproblemalsoexistsormaybemoreseriousinChina.InChina,theworkofevacuationandpedestriandatacollectionandstudyfromexperimentsandrealeventsislessthantheEuropeanandAmericancountries.ComputersimulationstudiesonevacuationinChinaoftenrelyonforeigndata.Evacuationbehaviourduringevacuationhasstrongdependenceonhumancharacteristics,suchasphysicalfeature,culturalbackgrounds,habitsandemergencytraining.Obviously,therearemustbesomedifferencesbetweenChineseandforeignerforevacuationbehaviour.UsingforeigndatatostudyevacuationofChinesecrowdwillinevitablygreatlyreducetheaccuracyandcredibilityofcomputersimulationofevacuation.So,itisnecessarytoconductresearchworkondatacollectionandanalysisofevacuationbehaviourinChina.

Inthismanuscript,datacollectionandanalysisofevacuationbehaviouriscarriedoutbymeansofquestionnairesurvey,evacuationdrillandcasestudy.Wefocusonhumanbehaviourbeforeevacuationthroughquestionnairesurveyandvideoinformation,andpre-movementtimeisdiscussedin-depth.Threemainparametersareabstractedinthisstudy,namelyresponsetime,firstactionandpre-movementtime.Theabstractionrulesareasfollows:Responsetimemeansthetimethattheoccupantspendsinrealizingtheemergencyandbeforehe/sheconductsthefirstaction.Firstreactionisthereactionthattheoccupantconductsafterhe/sherealizetheemergency.Pre-movementtime,thetimethat after the emergency occurs and before the occupant choose the right evacuation route.

Weconductsixunannouncedevacuationdrillsinameetingroom.Therewereabout60peopleinthemeetingroomwhenweconductthedrill.Acertainsmokematerialwasusedtosimulatesmokeoffire,whichwillproducecoldandharmlesssmoke.Somefirenoiseisproducedfromaudioequipmentinthedeviceroom.Duringthedrill,allthepeopleinthemeetingroomwouldseethe‘fire’smokeandhearthesoundof‘fire’,buttheywouldn’tdirectlyfacethefire.Threeevacuationdrillsareusedtodiscussthehumanbehaviourbeforeevacuation.Theotherthreeevacuationdrillsarenotverysuccessfulduetotheinfluenceofstaff.Also,questionnairesurveyisconductedtomakeaqualitativeanalysisofevacuationbehaviour.Analyzedandcomparedwithdatafromthreeevacuationdrills,itcomestoresultsasfollows.Firstly,peopleindifferentgroupsresponddifferently,whichresultsinsomedifferenceofpre-movementtime.Secondly,accordingtodatafromthevideoinformation,bimodaldistributionofpre-movementtimeisgotten.Analysisbygenderidentity,pre-movementoffemaleislessthanthatofmale,

whichcouldbethereasonofbimodaldistributionofpre-movement.

Twotypicalentertainmentplacesfirecasesareusedtostudythehumanbehaviourduringfire.Oncefire,personneleffectiveevacuationwouldbeunderseriousthreatandresultinfirecasualtyaccidentsintypicalentertainmentplacesforitscomplexdecorationandhigh-densityofpersonnelduringbusiness.Wefortunatelygotthebuildingstructureandvideodataofthesetwofirecase.Bothfirecasesresultfromfireworks,whichignitedtheceilingdecoratedofsound-absorbingfoamandotherflammableandtoxicmaterials.Bythevideoinformation,wedothesameanalysisandstudyasunannouncedevacuationdrills.Wecanobserveabout100people(can’tdistinguishgender)inonecaseand40people(29maleand11female)inanother.Allthepeopledidn’tevacuateimmediatelyuntiltheyrealizedthedangeroffire.Thebimodaldistributionofpre-movementtimeisalsogottenfromthetworealfirecases,althoughthemeanpre-movementtimeislessthanthatofunannouncedevacuationdrills.Thefasterbeginningofevacuationinrealfirecaseresultsfromtherapidfiredevelopmentintheentertainmentplaces,whichdecoratedfastburningmaterial.

Fromthestudyofunannouncedevacuationdrillsandrealfirecases,wegotdifferentmeanpre-movementtime,whichareabout18secondsinrealfirecasesandabout90secondsinunannouncedevacuationdrills.Theeffectofhightemperaturefieldandharmfulsmokeinrealfirewillenhancepeople‘sconsciousnessoffiredanger.Althoughpre-movementtimeisdifferentbetweenunannouncedevacuationdrillsandrealfirecases,itshowsthesamebimodaldistribution,whichisdifferentfromthetraditionalnormaldistributionandlog-normaldistribution.Thisfindingwillhelpfulforcomputersimulationofevacuation,especiallyforthosewhosimulatewholeprocessofevacuation.

Modeling pedestrian route choice during large public gatherings

Elise Miller-Hooks, University of Maryland, College Park MD USALei Feng, University of Maryland, College Park MD USA

Effectivemanagementofcrowdmovementisneededinlargepublicgatheringstoensureefficientandsafeingressandegresstoandfromtheeventorinthecaseofevacuationbyfoot.Suchgatheringsariseforavarietyofpurposesandmaybeheldinamyriadofvenues,includingforexample,complexbuildings,publictransportationstations,sportsstadiumsandcoliseums,amusementparks,commercialshoppingfacilities,andparklands,amongothers.Insuchgatherings,crowdsaredirectedthroughpassagewaysandopenareasdesignedforthepurposeofaccessingtheevent.Thephysicallayoutofthesepassagewaysprovidesasetofoptionsfromwhichpedestrianscanchoose.Theseoptionsmustbedesignedtosupportlargevolumesofpeople.Thespeedwithwhichapedestrianwillmovethroughthepassagewaydependsonitsphysicaldesignandthenumberofotherpedestriansutilizingitatthesametime.Thetimeforingressoregresstoorfromtheeventdependsontheseriesofchoicesthepedestrianmakesinnavigatingthephysicallayoutandtheinteractionswithotherpedestrianswithsimilarorcompetinggoals.Thispaperdescribesanetworkoptimization-basedmodelingandsolutionframeworkforassessingpedestrianresponsetothephysicallayoutofavenue’singressandegressroutes during such large public gatherings.

Thephysicaldesignofafacilityislimitedbyitspermanentinfrastructure.However,moveablebarriers,gatesandotherdevicescanbeusedtoorganizethespacewithinthatinfrastructure.Howthatspaceisorganizedaffectsthespeedwithwhichpedestrianswithinacrowdcanmove,theshapethecrowdcanform(e.g.singlefile),availablerouteoptionsandcircuity.Inmanycases,itmustalsofacilitateopposingflows for those pedestrians with dissimilar goals.

Giventhatthepedestrianshaveinformationabouttheirrouteoptions,howeachindividualchoosesaparticularroutedependsonhisorherpreferences.Thatis,eachindividualassignshisorherownvaluetodifferentcharacteristicsoftherouteoptions.Thisvaluesystemcanbecapturedinautilityfunction.Withinthecontextofthepublicgathering,eachindividualchoosestheroutetotheeventoroutoftheevent with the maximum utility.

Theconceptofroutechoiceinvehiculartrafficflowiswelldeveloped.Pedestrians,however,havemoredegreesoffreedominmovementandoftenchoosetotravelingroups.Suchgroupsariseinvehiculartrafficscenarios,butthesegroupsaretypicallyhousedwithinasinglevehicle.Forexample,afamilywilltravelwithinthesamecarorlargergroupswilltravelinabus.Thesegroups,thus,willneverbefacedwiththepossibilityofbeingsplitapart.Otherswhoseektoaccessthevenuetogetherbutindifferentvehicleswilloftenneedtobewillingtomeetatthedestination.Inthecontextofpedestrianmovement,however,groupsmustmakeaconcertedefforttomovetogetherandnotbesplitapart.Forexample,parentswillnotwishtobeseparatedfromtheirchildren.Thus,whileeachpersonwithinthe

familyisanindividual(i.e.aunitofflow)andisfreetomakehisorherowndecisionsinresponsetodirectivesfromcrowdmanagersorthephysicallayout,anyeffectivecrowdmanagementplanmustfacilitatethemovementofallmembersofthefamilyasagroup.Thatis,thegroupmustbepermittedtostay together and accommodations must be made to support this group movement. In mathematically modelingtheflowofpedestriansthroughaphysicallayout,flows,thus,mustconsistofnotonlyindividualmovements,butlargermovementsinvolvinggroupsofindividuals.

Theproposednetworkmodelingandsolutionframeworksupportthemovementofbothindividualsandgroups.Theapproachexploitsconceptsofutilitymaximizationandrecognizesthattheutilityofaroutedependsonbothfixedroutecharacteristics,suchaslengthorgrade,andcharacteristicsthatdependonthechoicesmadebyotherswhosimultaneouslyseekpassagealongthesameroutesaffectingthetraveltimealongtheroutesforallusers.Theoverallproblemofestimatingwhichroutesthepedestrianswilltakegiventheoptions,whichareboundedbythephysicalenvironment,andthefeaturesoftheoptionscharacterizedbytheirutilities,isknownasatrafficassignmentproblem,andistermed a pedestrian assignment problem in this context.

Assignmentproblemsforvehiculartraffichavereceivedenormousattentionintheliterature.Becausegroupbehaviorisnotconsideredinvehiculartransport,thedevelopedmodelsandalgorithmsfortrafficassignmentcannotbeapplieddirectlyinthemovementofpedestrianswheregroupbehaviormustbeconsidered.Onereasonforthisisthatthemarginalimpactofthedecisionofoneflowunitin pedestrian assignment where group behavior is modeled must account for the impact of group size.Moreover,pedestriansaresensitivetodifferentcharacteristicsofthephysicalenvironmentincomparisontodriverswithmotorizedmeansofmovement.Forexample,inpedestrianmovement,thelevelofphysicalexertionrequiredtotravel,whetherupaflightofstairs,alongaramporbyescalator,shouldbetakenintoaccount.Inthispaper,autilityfunctionisproposedthatincorporatesthoseelements of the route that will impact pedestrian choice.

Usingthisconceptofutilitymaximization,thepedestrianroutechoiceproblemcanbemodeledasann-playergame,whereeachplayerisagroup(individualsaremodeledasgroupsofone),eachwithitsownutilityfunction.Apure-strategyNashequilibriumofthegameissought.Thisdiffersfromroutechoicemodelsappliedinthecontextofvehiculartraffic,whichseekauserequilibriumorsimilarstochasticuserequilibrium,approachesthatcannotcapturegroupdecisions.ThisNashequilibriumapproachwillbepresentedinthepaper,andresultsofnumericalexperimentswillbeprovidedtodemonstrateitseffectivenessinmodelingpedestrianresponsetoavenue’sphysicallayout.

Simulation of city evacuation coupled to flood dynamics

Alexander Mordvintsev, National Research University of Information Technologies, Mechanics and Optics, St. Petersberg RUSSIAValeria Krzhizhanovskaya, University of Amsterdam, Amsterdam NETHERLANDSMike Lees, Nanyang Technological University, Singapore SINGAPOREPeter Sloot, University of Amsterdam, Amsterdam NETHERLANDS

Crowdmodellingisoneofthekeycomponentsofriskanalysisandevacuationplanninginemergencysituations.Inthispaperwepresentasimulationenvironmentforexperimentingwithdifferentcityevacuationscenarios.Inparticularweconsiderthecaseofcityflooding.Thesimulationcouplesafloodmodelwithacrowdescapemodelforscenariosoftwocityregions:inSt.Petersburg,RussiaandinAmsterdam,Netherlands.

Themodelconsistsoftwomainparts:hydrodynamicfloodingmodelandcrowdescapemodel.Weusearapidflood-spreadingmodel[1]forfloodwaterpropagation.Themodelreceiveswaterflowratesdischarged into floodplain areas from breached or overtopped defences and then spreads the water over thefloodplainaccordingtothecitytopography.Theprimaryoutputsofthemodelarewaterlevelsandflow velocities in the area.

Thedevelopedagent-basedcrowdmodelmimicsthebehaviourofpedestriansheadingfromdangerousregionstowardssafeareas.Itusestheoutputofthefloodsimulationtodrivetheevacuationprocess,tocomputeavailableevacuationpathsandtotracktheagentstrappedintheflood.Inputsofthemodelare:

- Citytopographydata,includingrasterheightmap(alsousedforfloodmodelling),andvector obstaclemap,whichcontainslocationsofbuildings,fencesandotherobstacles.Somebuildings ormapareasaremarkedasSafeHavens.- Demographicsdata,includingpopulationdistributionofagentsacrossthemodelledarea,the distributionofwalkingspeedsandotheragentcharacteristics.- Resultsofhydrodynamicmodellingofthefloodthatconsistsofwaterdepthdynamics.

Thestateofeveryagentisdefinedbyitsspatialposition,dangerawareness,maximalspeed,radius(approximating body projection on the ground plane) and the maximal water depth that the agent cantraverse.Themodelalsoincludessomesocialrelationshipsinourmodel.Forexample,somebehaviours that cause families to stay together (family members follow the leader who plans the evacuationpathandwaitsforothermembersincaseofwalkingspeeddifference).

Thecrowdmodeltakesthefollowingactionsateachtimestepofthesimulation:

1.Acquiresthefloodsimulationresultatcurrenttimestepandupdatestheobstaclemap.Notethatthefloodandcrowdmodelsmayusedifferenttimesteps.

2.Next,aglobalpathmaptothesafetyzonesiscalculated,thisisintheformofacontinuousdirectionfieldwhichconsidersthecurrentwaterdepthandobstaclelocations.WeuseamodifiedversionofDijkstra’salgorithm[2,3]onaCartesiangrid.Theagentsfollowapre-computeddirectionateverygridcell,thusreachingthesafezonebyanear-optimalroute.

3.Agentdesiredvelocityvectorsarethenupdated.Informedagentsmovealongthepre-computedroute.Uninformedagentsmoverandomlyorstayinoneplace.Socialrelationships(followingtheleader,waitingforfamilymembers,etc.)areincorporatedatthisstage.

4.WeusetheRVO2library[4]toimplementagent-agentandagent-obstaclecollisionavoidancelogic.Thelibraryusesanobstaclemaptocomputeagentvelocityvectorsavoidingcollisions.Herewewillalsoconsiderthecontinuousmechanicscollisionhandlingmethods[3,5,6].

5.Inthefinalstepswemoveagentsaccordingtotheirvelocitiesandupdatetheirknowledge-state.Agentswitnessingthewaterlevelriseareinformedandcommunicateawarningaroundthem.Agentstrapped in areas with too high water are considered drowned.

Resultsoftheproposedmodelareagentpositions,velocitiesandstatusesateachtimestep.Thesedataareusedtoanalyzejunctions,safeshelters,andevacuationeffectivenessindifferentfloodscenarios.

Ourmodelusesaglobalpathmapforagentpathplanning.Weassumethatallagentsknowthecitymapandflooddirection.Suchanidealsituationispossiblewithaninstantinformationtransferbetween the agents.

Ourcurrentimplementationissimulatingfiftythousandagentsinanareaofafewsquarekilometersseveraltimesfasterthanreal-timeonadesktopcomputer.ThecrowdsimulationpipelineisimplementedusingPythonandNumPy.C++codeisdevelopedforcomputationallyintensivepartsofthecode,suchaspathplanning.WealsouseanOpenMPparallelizedversionofRVO2[4].Withthatwehavecreatedasimulationenvironmentsuitableforexperimentswithdifferentscenariosofcityfloodingandcrowdevacuationstrategies.WehavesimulatedevacuationofVasilyevskyIslandinSt.PetersburgandScienceParkinAmsterdam.Furtherweplantoevaluatetheefficiencyofdifferentcontrol strategies and to experiment with distributed control through individual information devices suchassmart-phones.

ACKNOWLODGEMENTThisworkissupportedbytheLeadingScientistProgramoftheRussianFederation,contract11.G34.31.0019;bytheEUFP7projectUrbanFlood,grantN248767;andsponsoredbytheBiGGridprojectBG-020-10#2010/01550/NCFwithfinancialsupportfromtheNetherlandsOrganisationforScientificResearchNWO.

REFERENCES

[1] B. Gouldby et al. A methodology for regional-scale flood risk assessment. Proceedings of the Institution of Civil Engineers, 2008.

[2] J.N. Tsitsiklis. Efficient algorithms for globally optimal trajectories. IEEE Transactions on Automatic Control, Sep. 1995.

[3] A. Treuille, S. Cooper, Z. Popović. Continuum Crowds. ACM Transactions on Graphics, SIGGRAPH, 2006.

[4] J. van den Berg, D. Ming Lin Manocha. Reciprocal Velocity Obstacles for real-time multi-agent navigation. ICRA, 2008.

[5] Narain R. et al. Aggregate Dynamics for Dense Crowd Simulation. SIGGRAPH Asia, 2009.

[6] J. Shopf et al. GPU Crowd Simulation. SIGGRAPH Asia, 2008.

Poster exhibition

Unsorted list

Abdulaziz Aljohani, University of Birmingham, Birmingham UNITED KINGDOM Pedestrian crowd dynamics during pilgrimage

Maik Boltes, Forschungszentrum Jülich GmbH, Jülich GERMANY Tracking people in crowded scenes

Angelika Kneidl, Technische Universität München, München GERMANY Using a multi-scale model for simulating pedestrian behavior

S.B. Liu, City University of Hong Kong, Hong Kong CHINA On the simulation for rail tunnel evacuation with cross-passageways

Kristian Schatz, Technische Universität Darmstadt, Darmstadt GERMANY Investigating human factors in fire evacuation – a serious-gaming approach

Enrico Ronchi, Politecnico di Bari, Bari ITALY Validation and calibration of the EXIT89 evacuation model for road tunnel evacuation applications

Pedestrian crowd dynamics during pilgrimage

Abdulaziz Aljohani, University of Birmingham, Birmingham UNITED KINGDOM

ABSTRACTMorethanthreemillionpilgrimsannuallygatheringinMakkah,SaudiArabia,PilgrimsmovementsduringtheHajjisacomplexphenomenonwhichcanbedefinedasasetofprescribedmovementsofdifferentgroupsofpilgrimsandindividualsinthecourseofperformingthereligiousritualsoftheHajjbetweenMakkahandtheholyareasatspecifichoursonthe8thto12thdaysofthepilgrimagemonth.TheHajjseasonhaswitnessedseveralfatalincidents.In2006,340pilgrimsdiedinastampedeofpilgrimsatthestageofthestoningritualattheJamaratinMina,Onemaincauseisthecross-trafficinahighdensityarea.Thisstudyinvestigatescharacteristicsofpedestrianflowduringpilgrimagetime.DatawerecollectedinoneofthehighdensityStreets,theAjyadStreet,Makkah,SaudiArabia.Theaimsaretoconducttheflowstudy,andtoexaminetherelationshipsofpedestriansspeed,flowanddensity

GENERALAPPROACHPilgrimsmovementsduringtheHajjisacomplexphenomenonwhichcanbedefinedasasetofprescribedmovementsofdifferentgroupsofpilgrimsandindividualsinthecourseofperformingthereligiousritualsoftheHajjbetweenMakkahandtheholyareasatspecifichoursonthe8thto12thdaysofthepilgrimagemonthaddEuropeancalendartoo(Dhul-Hijja).Onthemorningofthe10thdayofthepilgrimagemonth,mostpilgrimscometoMakkahfromMinawheretheyaresupposedtoremainatMinafromthe10thtothe12thorstayonemoreday-the13thofDhul-Hijja-tostonethedevil(Jamarat).Duringeachofthesedays,pilgrimscometotheHolyMosqueatMakkaheithertoperformTawafortopray.Thesedays,theHajjseasonhaswitnessedseveralfatalincidents.In2006,340pilgrimsdiedinastampedeofpilgrimsatthestageofthestoningritualattheJamaratinMina,in2004,251pilgrimsdiedinastampedeunderthefeetofpilgrimsattheJamarat.In2003,14peoplewerecrushedtodeathwhenpilgrimsreturningfromJamarat.In2001,35pilgrimswerekilledinthecrowdattheJamarat.In1998,118pilgrimsdiedpilgrimscrushedtodeathaftersufferingapanicsituationontheimpactofthefallofsomepeopleduringthestoningritual.In1997,atleast340pilgrimsdieafterafireinthepilgrimstentsinMina.In1994,270pilgrimswerekilledinastampededuringthestoningritual.In1990,1,426pilgrimswerekilledinastampedeinanovercrowdedtunnelleadingtotheHolyMosqueinMecca.(SaudiCivilDefence,2006).Oneprimecauseiscross-trafficinahighlydensearea.Pilgrimsfinishingtheactivityleave,orfighttheirwayout,whilearrivalstothesite,enter,orfighttheirwayin.Thistypeoftrafficcreatesanunpredictablemovementpatternwhicheasilyresultsinfalling,causingbodilyinjuriesorfatalities.AttheseasonsofUmrah-Visiting,morethantwomillionpersonscometoMakkahandMadinah,withinarateof500.000personseverymonth.InthemonthofRamadan,thisnumberincreasestoaroundthreemillionpilgrimsinMakkah.Theincreasingnumberofworshippersthatcomeatthesametime,andinaspecificgeographicalarea,increasesthechancesoflarge-scale

disasters.InvestigationofHajjCrowdinAjyadStreet(themainpedestrianpilgrims’walkwaytotheHolyMosque)duringHajjtimeconsideringthepedestriancharacteristicsaccordingtoraiseandethnicbackground.Thishasdifferentcrowddynamicsparameters;suchasinputcapacity,speed,flowanddensity.TheinputdataisobtainedbyimageprocessingatinletareaatdifferenttimesduringthemostpopulateddayduringtheHajjinAjyadStreet.TheaforementionedparametersaresimulatedbytheSimWalksimulationsoftwarewhichyieldsthepilgrimscrowdcharacteristicsalongthepathofAjyadStreet.Thesimulatedresultsarecontrastedwithafieldmeasureddataatselectivepointsalongthepath.Thefieldvideodataatselectivehoursarechosenandthevideoaredecomposedtosuccessiveimages,andimageprocessingalgorithmhasbeenbuilttorecogniseandtrackeachindividualpilgrimalongthecourseofmotion.Thesefieldextracteddataareusedtovalidatethesimulatedresultsintermsofpedestrianflowrate,speedanddensity.ThisinvestigationcanbeconsideredasabuildingblockforacomprehensivesimulationcodeincludingallHajjactivitiesduringthewholeperiodofHajj.Thisgeneralsimulationcouldbelinkedtocameraslocatedatselectiveplacesbywhichanimageprocessingcodewillinterpretthoseimagesintoaninputdatatothegeneralsimulationprogram.Thesimulationresultswillenabletheauthoritiestodiagnose,predictoravoidpilgrims’crowdproblemsIt is anticipated an overall calibration and validation of this general code is necessary to improve the decisionmakingprocess.

DATACOLLECTIONPLANThreedatacollectionexerciseswereperformedinordertocollectthenecessaryinformationandsupportanin-depthinvestigationofthepilgrimcrowddynamicscharacteristics.Forthispurposethefollowing factors were determined:•Pilgrims`flowrate,speed,anddensityatAjyadStreet;oneofthemainroadsleadingtotheHolyMosque•Movementpatternsofpilgrimse.g.unidirectional,bi-directional,andmulti-directionalpedestrianflow.•Pilgrims’densityatthePlazaoftheHolyMosque;•HighdensityspotsinfrontofthemaingatesoftheHolyMosque;•Pilgrimsdemographiccharacteristicswhichaffecttheirmovement.

Inthefirstphase(2008)adiagnosticmeasurementsofcrowdduringHajjtimehasbeencollectedusingstillcamera.Thistohelpofdeterminethemostcriticalplaceswhicheffectthecrowdynamics.Inthelightofthisprimarydatacollection,AjyadstreetandtheHolyMosquePlazahavechosentothecasestudyinthepresentwork.Itisanticipatedthatanybetterunderstandingofthecrowddynamicswillalleviate the problem of overcrowding

DATACOLLECTIONMETHODESBothquantitativeandqualitativeresearchmethodswereusedinordertostudy,analyse,anddiscussthepilgrimscrowddynamicsatAjyadstreetandtheHolyMosquePlazaduringtheHajj.Acombinationofthesemethodsoffersfullopportunitiestostudythepilgrimcrowdmovementsfromdifferentperspectives.Thesemethodsinclude:•VideoandstillimagecamerasusedtocollectthepilgrimscrowddynamicsfromAjyadStreetandtheHolyMosquePlaza.•Visualinspectionofthevideorecordingsandhighresolutionstillimagestoobtainbetterunderstandingofthegeneralmechanismsofhowpilgrimsinteract,(forinstance,howtheytendtoover-takeeachother,organisethemselvesintolanesandphysicallyinteractwitheachother)aswellaspilgrimscrowdtrajectories,movementtypes,andovercrowdingspots.•Extractingthefundamentalpilgrimdynamicscharacteristics;speed,densityandflow.bymanually

counting. •ObservationofcrowdmovementpatternsduringpilgrimageatAjyadstreet,andPlazaoftheHolyMosque.•Statisticalanalysisofpilgrim’sdemographiccharacteristics,e.g.gender,age,andethnic.•ComputersimulationofpilgrimscrowddynamicsduringtheHajjseason.•Validationofpilgrimcrowddynamicsimulation.

Thepresentdataareobtainedmanuallyfromallvideorecordings;itproposedtoprocessallvideorecordings. Footages will be processed a software to obtained the crowd dynamics characteristics at theselectedmeasuringcitesofAjyadstreet.Thesoftwareisbasedondecomposingeachvideodataintotheirconsequentialframes.Foreachframe,thesoftwarerecognizeseachindividualpilgrimbydetectionofitsedgesorboundaries.Acountingalgorithmisusedtogetthecrowddensityatthetimeofeachframe.Further,imageprocessingofconsecutiveframeswillyieldthemovementspeedforeachindividualalongwiththecrowdbulkmovement.Aselectedgroupofindividual(pilgrim)willbechosenasacasestudytoinvestigatethemovementcharacteristicswithrespecttoplace,timeanddayof event.

Tracking people in crowded scenes

Maik Boltes, Forschungszentrum Jülich GmbH, Jülich GERMANYArmin Seyfried, Forschungszentrum Jülich GmbH, Jülich GERMANY

Fortheproperunderstandingandmodellingofpedestriandynamics,reliableempiricaldataisnecessaryforanalysisandverification.Collectingthetrajectoriesofeverypersonwithahightemporalandspatialresolutionallowsadetailedanalysisofmovementandthecalibrationandverificationofmicroscopicmodelsinspaceandtime[Steffen2011,Chraibi2010].

Inrecentyearswehaveperformedanextensiveseriesofwell-definedexperimentswithupto350peopletostudythemovementofpedestriansindifferentsituations[Seyfried2010,Holl2009].Theselaboratory experiments give us the opportunity to analyse parameters of interest under controlled conditions.Thevariabilityallowsasurveyofaparameterrangee.g.forthebottleneckwidthorlength,orthedensityinsideacorridor.Parameters,likethedensity,canbesettovaluesseldomseeninfieldstudies (e.g. very high densities). For such experiments the characteristics of the test persons (e.g. culture,fitness,age,gender,bodyheight)canbedetermined.

For the analysis of these experiments we have developed a software to automatically extract trajectoriesfromvideorecordingsofmarkedpeopleonplaneground[Boltes2010]anduneventerrain[Boltes2011].Theprogramisabletohandlelensdistortionandhighpedestriandensities.Forexperiments e.g. at stairs but also for experiments on plane ground stereo recordings are needed to get spatialtrajectoriesandtotaketheperspectivedistortionintoaccount.

Despitetheabovementionedadvantagestheexperimentsunderlaboratoryconditionshavealsodrawbacks.Thenumberofexperimentsislimitedduetothecostsofthetestpersonsandforbuildingtheartificialenvironments.Thusthevarianceofthestudiedparameterislimited.Alsothecombinationofdifferencesinsideadetectedgroupcannotbecoveredbylaboratoryexperiments.

Thereforewepresentanewapproachtodetectpedestrianswithoutmarkeralsoincrowdedscenestofacilitatefieldstudiesandtheeasierrealizationofmoderatedexperimentsinrealenvironments.Thenewlyintroducedmethodbasedontheanalysisofthedepthfieldofstereorecordingstakenfromoverheadofthepedestrians.Theoverheadrecordingsperpendiculartothefloorallowaviewwithoutocclusionforarangeofbodyheights,sothatamicroscopicdetectionandtrackingwithoutestimationofthepersons‘routecanbeperformed.

Therehasalreadybeendonealotofworkinthefieldofpedestriandetection.Mostoftheapproachesareformonocularcamerasandslantedviewslikefromsurveillancecameras,anddensitieswhichresultonlyintemporaryorpartlyocclusion.Oneofthebestresultsforthesescenarioscanbefoundin[Schwartz2009].Existingtechniquesfortrajectoryextractionforstereocameras,suchas

[Harville2004],dependonaccuratesegmentationofforegroundobjects.Fordensecrowdssuchasinour experiments these methods are not be applicable or could only detect groups of people.

Ourextractionmethodcopesalsowithcrowdedscenes.Itdirectlyusestheperspectivedepthfield,anddoesnotuselaboriousplan-viewstatisticstospeedupandsimplifytheextractionstep.Thedepthfieldcontainsthedistancetothecameraforeverypixelandisinverselyproportionaltothedisparitymap,whichdescribesthepixeloffsetofbothcameraviewfieldsofthestereocameraforeverypixel.

Thenewmethodworksnowasfollows.Theidentificationofthepeopleisdoneonlyusingtheshapeof the top part of their body especially the head and shoulders. If we want to identify people only bytheirshape,abackgroundsubtractionhastobeperformedbeforetoreducethenumberoffalsepositivedetections.Pixelsarepartofthebackgroundandthusareignoredinthedetectionprocess,ifthedistancetoaperspectivedepthfieldofthebackgroundissmallerthanathresholdvalue.Theperspectivedepthfieldofthebackgroundissetoncewiththescenedesertedorissettoacautiouslyadapted maximum distance during all frames.

Toidentifypedestriansbymeansofthedepthfieldwedeterminedirectedisolinesofthesamedistancetothecameraatequidistantdepthlevelsfortheupperbodypart.Inadvancethedepthfieldisadaptedbyreplacingvaluescoveredbythebackgroundmaskwiththefurthestvaluebelongingtotheforeground.

Theisolinesenclosingaminimumandmaximumofpixelandwithasmallratiobetweenthelengthof the isoline and the enclosed area (to eliminate isolines with big dents) are approximated by ellipses. Theellipsesallowaneasieraccesstotheglobalshape.Byscanningthedepthfieldfromtheheaddownwardsapyramidofellipsesfortheupperbodypartofeverypersonisbuildup.Thesemeasuredpeoples’pyramidalellipsesstacks(PES)arematchedagainstavarianceofpeoplemodelswheretheperspectiveviewhastotakeintoaccount.ThePESarecomparedtoellipsesstackswegenerateofsynthetic models by raytracing a virtual scene simulating the depth sensing of a stereo camera.

ThePESresultingfromameasureddisparitymapareunstableandthusthecentreofthetopmostellipseisnotagoodpointforthecentreofthehead.TostabilizetheprocedureandthustogetsmoothertrajectoriesweutilizethedistortedaxisofthePESforamoresettledcentre.Thesmoothnessoftrajectoriesresultingfrompeopledetectionwithandwithoutmarkeriscomparedinthepaper,because it is important for further analysis (e.g. instant velocity) of the trajectory data. AstrictrejectionofnotproperfittingPESavoidsfalsedetectionssinceitisnotnecessaryfortrackingtodetectapersoneveryframe.FortrackingthepedestrianstherobustpyramidaliterativeLucasKanadefeaturetrackerisusedtojointhesamedetectedpedestrianinsuccessiveframesorbypassoverframeswhereaspecificpedestriancouldnotbelocated.

Thetrackingresultsexceedtheresultsofallformermethodsfortrackingofmarkerlesspedestrians.Besidesthis,themarkerlessdetectioncanimprovetherobustnessofthemarkerbaseddetection,asdetectedmarkernotlyingonanelevationdescribingapersons’headcanberejected.

REFERENCES

[Boltes2010] M. Boltes, A. Seyfried, B. Steffen, and A. Schadschneider. Automatic Extraction of Pedestrian Trajectories from Video Recordings. In: [PED2008], 43–54.

[Boltes2011] M. Boltes, A. Seyfried, B. Steffen, and A. Schadschneider. Using Stereo Recordings to Extract Pedestrian Trajectories Automatically in Space. In: [PED2010], 751–754.

[Chraibi2010] M. Chraibi, A. Seyfried, and A. Schadschneider. Generalized centrifugal force model for pedestrian dynamics. Physical Review E, 82:046111, 2010.

[Holl2009] S. Holl and A. Seyfried. Hermes - an Evacuation Assistant for Mass Events. inSiDe, 7(1): 60–61, 2009.

[Harville2004] M. Harville. Stereo person tracking with adaptive plan-view templates of height and occupancy statistics. Image and Vision Computing 22, Statistical Methods in Video Processing, 127 – 142, 2004.

[PED2008] W. W. F. Klingsch, C. Rogsch, A. Schadschneider, and M. Schreckenberg (ed.), Pedestrian and Evacuation Dynamics 2008, Springer, 2010.

[PED2010] R. D. Peacock, E. D. Kuligowski, and J. D. Averill (ed.), Pedestrian and Evacuation Dynamics 2010, Springer, 2011.

[Schwartz2009] W. R. Schwartz, A. Kembhavi, D. Harwood, and L. S. Davis. Human Detection Using Partial Least Squares Analysis. International Conference on Computer Vision (ICCV 2009), Kyoto, Japan, 2009.

[Seyfried2010] A. Seyfried, M. Boltes, J. Kähler, W. Klingsch, A. Portz, T. Rupprecht, A. Schadschneider, B. Steffen, and A. Winkens. Enhanced empirical data for the fundamental diagram and the flow through bottlenecks. In: [PED2008], 145-156.

[Steffen2011] B. Steffen, M. Boltes, and A. Seyfried. Improved Methods for Checking Forces Based Models of Pedestrian Dynamics. In: [PED2010], 885–888.

Using a multi-scale model for simulating pedestrian behavior

Angelika Kneidl, Technische Universität München, München GERMANYDirk Hartmann, Siemens AG, München GERMANYAndré Borrmann, Technische Universität München, München GERMANY

Simulationofcrowddynamicshasbecomeanimportantfieldofresearchinthelastyears.Avarietyofdifferentapproacheshavebeendevelopedaccordingtodifferentobjectives.Theseapproachesmodelcrowdbehaviorondifferentscales:Smallscalemodels,likecellularautomata(Blue&Adler1999)orsocialforcemodels(Helbing&Molnár1995),modeleachindividualandmovetheseindividualstoasetofrules,untilallindividualshavereacheddesignateddestination..Onthesmallscaleobjectivepedestrians typically steer directly towards a visible target trying to avoid other pedestrians. Secondly,medium-scalemodelsconsiderslightlymorecomplexnavigationbehaviortypicallynotsteeringdirectlytowardsadestination,e.g.ifthedestinationliesbehindanobstacle.Typically,thismediumscalenavigationbehavioriscombinedwithanappropriatesmallscalemodel.Asforthesmallscalemodelspedestrianbehaviorcanbemodeledeffectivelyusingpotentialsorforcesgivenasgradientsofpotentials.Bymovingalongthegradientofthesepotentialfields,pedestrianssteertowardsthedesireddestination,sinceitislocatedintheminimum.

Finally,thereexistlarge-scalemodels,whichconsistofnetworksorgraphs.Thesedatastructuresareusedtoapplyroutingalgorithmstoguidepedestriansthroughagraphornetworkinordertoreachacertaindestination.Anexampleforsuchmodelsisanetworkoptimizationmodel(e.g.Hamacher&Tjandra2002).

Inthiscontribution,acombinationofallthreedifferentscalesisdiscussed.Webelievethatmodelingpedestrians’behavioriscomplexandneedstoconsiderdifferentlevelsofdetail.Inourmulti-scalemodel,themicroscopiclayerformsacellularautomaton,whichconsistsofhexagonalcells.Ineachtimestep,eachpedestriancanoccupyexactlyonecell.Byapplyingcertainupdaterulesineachupdatestep,whicharederivedfromthesmallscalelayer(avoidingotherpedestrians)andthemedium-scalelayer(navigationtowardsthedestination),eachpedestrianismovedtoafreeneighboringcell,whichislocatedclosertothedestination.Potentialsorso-callednavigationfieldsareusedtomodelthemedium-scalelayer.Byassigningrepellentforcestoobstaclesandothermovingpedestrians,anavigationfieldcanbegenerated,whichcontainstheshortestpathtothedestination.UsingtheFastMarchingMethod(Sethian1999),thesefieldscanbecreatedveryefficientlyontopofregulargrids–whichinthiscasearethedualgridsofthecellularautomaton.Thebasicideaistopropagateawave,whichslowsdownwhenpassingobstaclesandgetsfasteraroundthese.Now,pedestriansareabletowalkonthefastestpathfromthesourcetotheirdestination.Thiscombinationbetweensmall-scaleandmedium-scalelayershasbeenimplementedwithinmanysimulators,e.g.(Kretz2009),(Köster,Hartmann&Klein2010)or(Nishinarietal.2004).

Incontrasttotheassumptionstakeninthemodelsdescribedabove,notalwaystheshortestpathischosenbypedestriansconsideringreal-worldbehavior.Additionallytoconsideringothermovingpedestriansinthedirectvicinity(asadoptedbymanyexistingapproaches),otherpedestriansareconsideredaswellonthelarge-scalelayer.Thislayerisimplementedbyanavigationgraph,i.e.visibilitygraph,whichisderiveddirectlyfromthetopographyofagivenscenario(Kneidl,Borrmann&Hartmann2011).Usingthisgraph,firstofall,differenttypesofway-findingbehaviorslikefollowingthefastestpath,followingsignage,avoidcongestionscanbemodeledbyapplyingdifferentroutingalgorithms.Forallthesealgorithms,edgeweightshavetobedefinedinordertomeasuredifferentroutesthroughsuchagraph.Edgeweightderivationherebycanbedoneindifferentways:Thesimplestwouldbetotakethedistancebetweentwoadjacentvertices.Thisdoesnottakeintoaccountanyothermovingpedestrians.Asecondpossibilityistouseinsteadtraveltimes.Sincethevelocityofpedestriansdependsonthedensityofacrowd,thetraveltimeincreasesifanedgebecomesmorecrowded.

However,derivingdensitiesonedgesisambiguous.Steffen&Seyfried(2010)proposemethodstoderivelocaldensities.Withcellularautomata,pedestriansinneighboringcellscanbecountedefficiently.Still,thesemeasuresareestimates,sincethecorrespondingvelocityisanestimatedvalueitselfcorrespondingto(Weidmann1993).

Toovercometheseissues,weproposetousenavigationfields(sometimescalledfloorfields)fromthemedium-scalelayerandestimatefromthesetraveltimesoneachedge.Theconsiderednavigationfieldsaredynamicwithrespecttoexistingpedestrians,asproposedby(Hartmann2010),andareupdatedifacertaineventonanedgeoccurs.Insteadofhavingonlyonelargenavigationfieldforthewholescenario,whoseupdatewouldbecomputationallyveryintensiveifalargescenarioissimulated,wedefinemanysmallfields.Thesesmallfieldsareconstructedbyconnectingtwoadjacentverticesofthelargescalenavigationgraph,i.e.intermediatedestinations.Thisresultsindefiningonenavigationfieldforeachedge.Withthesesmallfields,theupdatecanbecalculatedveryefficiently.Inaddition,wecandirectlyretrievetheedgeweightbysimplytakingthevalueofthecellofthestartnodeofanedge,sinceweconstructthenavigationfieldpropagatingthewavefromtheendnodetothestartnode.Thevaluethusreferstothedetourapedestrianhastotakeinordertoevadeallothermovingpedestrians.Thisisamoreprecisemeasurefortheedgeweights,sinceitisdirectlyreadfromagivensituationandnotestimated through assumption as before.

Inourcontribution,wewilloutlinetheexactsetupofourmulti-layersimulationmodel,includingadetaileddescriptionofeachlayerandtheirinteraction.Theincreaseinrealismaswellinasincomputational performance will be underlined by ostensive examples.

REFERENCES

Blue, V and Adler, J 1999, ‚Cellular Automata Microsimulation of Bidirectional Pedestrian Flows‘, Transportation Research Record: Journal of the Transportation Research Board, vol. 1678, pp. 135–141.

Hamacher, HW and Tjandra, SA 2002, ‚Mathematical modelling of evacuation problems: A state of the art‘ in Pedestrian and Evacuation Dynamics. International Conference on Pedestrian and Evacuation Dynamics, held at the Gerhard-Mercator-University Duisburg, Germany, April 4 - 6, 2001; with 28 tables, ed M Schreckenberg, Springer, Berlin, pp. 227–266.

Hartmann, D 2010, ‚Adaptive pedestrian dynamics based on geodesics‘, New Journal of Physics, vol. 12, no. 4, p. 43032.

Helbing, D and Molnár, P 1995, ‚Social Force Model for Pedestrian Dynamics‘, Physical Review E, vol. 51, no. 5, pp. 4282–4286.

Kneidl, A, Borrmann, A and Hartmann, D 2011, ‚Generating sparse navigation graphs for microscopic pedestrian simulation models‘. Proceedings of the 2011 eg-ice Workshop, eds T Hartmann, P de Wilde and Y Rafiq.

Köster, G, Hartmann, D and Klein, W 2010, ‚Microscopic pedestrian simulations: From passenger exchange times to regional evacuation‘. Operations Research - Matering complexity.

Kretz, T 2009, ‚Pedestrian traffic: on the quickest path‘, J. Stat. Mech., vol. 2009, no. 03, pp. P03012. Nishinari, K, Kirchner, A, Namazi, A & Schadschneider, A 2004, ‚Extended floor field CA model for evacuation dynamics‘, IEICE Trans. Inf. Syst, E87D, pp. 726–732.

Sethian, JA 1999, Level Set Methods and Fast Marching Methods, Cambridge University Press.

Steffen, B and Seyfried, A 2010, ‚Methods for measuring pedestrian density, flow, speed and direction with minimal scatter‘, Physica A: Statistical and Theoretical Physics, vol. 389, no. 9, pp. 1902–1910.

Weidmann, U 1993, Transporttechnik der Fussgänger. Transporttechnische Eigenschaften des Fussgängerverkehrs (Literaturauswertung), ETH Zürich.

On the simulation for rail tunnel evacuation with cross-passageways

S.B. Liu, City University of Hong Kong, Hong Kong CHINAS.M. Lo, City University of Hong Kong, Hong Kong CHINAJian Ma, Southwest Jiaotong University, Chengdu CHINA

Accidentsinrailtunnelscannotbeseenveryoftenbycomparingwithdisasterssuchasbuildingfires.Butonceithappensthedamageisalwaysunpredictableandunacceptable.Ithasbeenreportedinthebook“Thehandbookoftunnelfiresafety”[1]thathundredsofserioustunnelfireshadcausedthelossoftensofthousandsoflivesduringthepastfewdecades.Tostudythenatureofevacuationfromrailtunnelsisessentialforreducingthehumanlifelostinarailtunnelaccident.Therehasalreadybeenanextensiveliteratureontheresearchoffiresafetyintunnels,butmostofthemhavebeenfocusedonthenatureoffireincidence,nottheevacuationprocess.Oswaldet.al.[2,3]havepresentedtheirstudiesabouttheevacuationintunnelfiresituationsbyexperimentsinthepastPEDconferences.Butsimulationworksonthistopicisstillveryrare.

ItisdenotedintheNFPA(NationalFireProtectionAssociation)130standard(StandardForFixedGuidewayTransitandPassengerRailSystems)[4]that“Thesystemshallincorporateawalksurfaceor other approved means for passengers to evacuate a train at any point along the trainway” and that “withinundergroundorenclosedtrainways,themaximumdistancebetweenexitsshallnotexceed762m”.Thencross-passagewaysorexitstairsareessentialforatunnellongerthan762m,especiallyfortwin tunnels. It also denotes that the means of egress within the trainway shall be provided with the minimumclearwidthof610mmatthewalkingsurfacelevelandcross-passagewayswithaminimumclearwidthof1120mmshallnotbefartherthan244mapart.Buttherealsituationsinengineeringprojectsaremorecomplexandsometimesneedfurtherperformance-basedevaluations.Inordertofigureouttheeffectsofdifferentdimensionsofthetunnelegresselements,tobuildareliablesimulationmodeleligiblefortunnelegresscaseswillcertainlybethefirstconcern.Butegressfromtunnelsisstillachallengingtopicconsideringthatthespecialnarrowandconfinedspace,thecomplexinteractionsbetweenpeopleandpeople,etc..Thevolumeofpublicationsforstudyingegressprocessfromrailtunnels,especiallythecomputersimulationstudies,islimited.

Inthepastmanyyears,largeamountofcomputersimulationmodelsforpedestrianandevacuationdynamicsinvariousenvironmentsandsituationshavebeendeveloped.Benefitingfromthedevelopmentcomputertechnology,microscopicmodelsarebecomingmoreandmorepopularrecently. Microscopic models are usually considered be divided into discrete models and continuous models.Discretemodels,suchastheCellularAutomaton(CA)model,theLatticeGas(LG)model,etc.usuallydiscretizethefloorplanintouniformrectangulargrids.Eachgridrepresentsacertaindimensionofspaceinreality,suchas0.5×0.5m2,andcaneitherbeoccupiedbyatmostonepersonorbevacant.Discretemodelsarepopularlyusedinmanyevacuationandpedestrianflowsimulationsoftwarebecauseofitshighcomputingefficiency.Butthefeasibilityofusingdiscretemodelsforthenarrowpassagesorveryconfinedareassuchasthewalkwayinsideatunnelshouldbefurther

examinedbecausethenarrowwalkwaysintunnelsaredifficulttobeexpressedby“grids”.Forexample,the610mmand710mmwidthwalkwaywillbethesameinadiscretemodelsincebothofthemwillbecome one gird width in the simulation.

Continuousmodelsusuallyemployforcesbetweenpeopleandpeople,peopleandenvironment,togoverningthemovementfromtheperspectiveofphysics.Thecontinuousmodelssufferfromlowcomputingefficiencyandarenoteasytobeappliedforpracticaluse,especiallyforcaseswherelargenumberofpeopleisinvolved.Buttheyallowpeopletomovecontinuouslyinthesimulationdomainin all the directions and thus their simulation results may be closer to reality in simulating detailed pedestrianmovingtrajectoriesandpatterns.Itisverylikelythatthecontinuousmodelscouldgetbetter results than discrete models on modeling cases with narrow passages or environments with very confinedarea(wherehowthespaceisrepresentedbecomesimportanttotheresults).Butreportonthistopic is very rare.

Inthisstudy,anagent-basedcontinuouscrowdsimulationmodelisdevelopedtostudythefeaturesoftunnelevacuation.The2Dfloorplanisdefinedbypointsandlinesandthehumanbodyisrepresentedasanagentwithmovementdirection,awarenessangleandawarenessdistance.Themodelisimplementedintwolevels:first,themacroscopiclevelforroutechoiceandroadmapnavigation;second,themicroscopiclevelforagentmovementandobstacleavoiding.Thegoverningrulesforagentmovementaredesignedbyconsideringseveralfactors,includingtheefficientofapproachingtheintermediatetargetposition,theagent’sobstacleavoidingbehavior,theinteractionstriggeredbythechangeofdistanceandmovingdirectionbetweendifferentagents.Themodelisabletomoveagentsinanydirectionincontinuousspace,thusenableustohandletheminorchangesoftunnelegresselements’dimensionssuchasa100mm’sincreaseofwalkwaywidth.

ByreferringtothesestandardsstatedinNFPA130,evacuationfromrailtunnelwithcross-passagewaysisstudiedbysimulationswiththeagentbasedmodel.Thisstudyisfocusingontheegresstimeandthepatternofthecrowdflowdynamicsunderdifferenttunnelegresselementdimensionsorscenarios.SimulationsfordifferentscenariosareconductedtoexaminetheinfluenceofthedifferentwalkwaywidthWandcross-passagewaysspacingSontheevacuationresults.Twoscenarios,scenarioAandBaresetuptostudytheinfluencesofWandS,correspondingly.InscenarioAthevalueofSisfixedandthevalueofWvariesfrom610mmto1610mm,whileinscenarioBthevalueofWisfixedandthevalueofSvariesfrom60mto294m.BothscenariosAandBaresimulatedbythetwokindsofmodels.Besides,giventhatdifferentemergencyalightingpositionsofthetrainmayleadtodifferentrelativelocationsofthetraindoorstothecross-passageways.Thus,differentevacuationstrategiesareexaminedanddiscussedbasedonthesimulationresultsinthisstudy.Theinteractionsbetweentheforeandafteragentswithinthequeueonthenarrowwalkwayandtheflowpatternsundersituationswithdifferentwalkwaydimensionsarealsointerestingproblemstobepresentedinthisstudy.

REFERENCES

[1] Carvel, R. and G. Marlair, A history of fire incidents in tunnels, in The handbook of tunnel fire safety, Alan Beard and R. Carvel, Editors. 2005, Thomas Telford: Heron Quay, London. p. 3-37.

[2] Oswald, M., H. Kirchberger, and C. Lebeda, Evacuation of a High Floor Metro Train in a Tunnel Situation: Experimental Findings, in Pedestrian and Evacuation Dynamics 2008. 2008. p. 67-81.

[3] Oswald, M., et al., Full-Scale Evacuation Experiments in a smoke filled Rail Carriage — a detailed study of passenger behaviour under reduced visibility, in Pedestrian and Evacuation Dynamics 2005. 2007. p. 41-55.

[4] NFPA, NFPA 130-Standard For Fixed Guideway Transit and Passenger Rail Systems 2010 Edition, in Chapter 6 Trainways. 2009: Batterymarch Park, Quincy, MA.

Investigating human factors in fire evacuation – a serious-gaming approach

Kristian Schatz, Technische Universität Darmstadt, Darmstadt GERMANYJosef Schlittenlacher, Technische Universität Darmstadt, Darmstadt GERMANYDaniel Ullrich, Technische Universität Darmstadt, Darmstadt GERMANYUwe Rüppel, Technische Universität Darmstadt, Darmstadt GERMANYWolfgang Ellermeier, Technische Universität Darmstadt, Darmstadt GERMANY

Anewwaytoreachfiresafetydesigndecisionsispresentlybeingestablishedasanalternativetotheuseofprescriptivecodes:theperformance-basedapproach.ThisprocessstartedafewyearsagodrivenbytheISO[1].Applyingtheperformance-basedapproachtofireprotectiondesignemphasizesthesafeperformanceofabuildingasawholeratherthanmeetingdetailedcoderequirements.Tothiseffect,firesafetyengineersmakeuseofcomputermodelsandsimulationstodescribetheexpectedspreadoffireandsmoke,andthesafetyevacuationitself[2].Sincetheprotectionofhumanlifeistheprimaryaimoftheperformance-basedapproach,thepredictingthebehaviorofpeopleindangerisanessentialpurposeofsuchmodeling.Particularly,therelevanthumanfactors(individualdecisionsandparameterstodescribehumanbehavior)havetobetakenintoaccount.AccordingtoSantosandAguirre[3],foranevacuationsimulation,threeanalyticaldimensionsneedtobeconsidered:thebuiltenvironment(physicallocation),themanagementofthisenvironment(signage,escaperoutes),andpsychologicalandsocial-organizationalcharacteristicsoftheoccupants.Tavares[4]pointedoutthatanevacuationsimulationmodelmustconsiderfourinteractions:occupants-structure;occupants-occupants;occupants-fire(incaseoffireevents)andfire-structure(forthispurpose,afiremodelshould be used).

Toexaminethesehumanfactorsdifferentdata-collectiontechniqueslikeinterviewswithsurvivors,onlinequestionnaires,mapexercisesorsimulationexperimentsareused.Thesemethodsoftencover only singular aspects of human behavior or studies are carried out independently from each other.Anotherpointisthatintervieweesmostlyknowthattheyarenotinadangeroussituationandthereforefeelnocognitiveemergencystress.However,intheFSEcommunitythereissomecontroversyaboutthispoint.SantosandAguirre[3]arguethatappropriatemethodsforvalidationofthehumanbehaviormodelarenotavailable.Sotheystatethatpropervalidationtoolshavetobedevelopedandthatthereforemultidisciplinarycollaborationisneeded.However,accordingtothem,theabilityto accurately and comprehensively simulate human behavior is missing from current computer evacuationmodels[3].AccordingtoKuligowskiandGwynne[5]thesemodelsoftenhavetheproblemthat the behavior simulated in the scenario is actually prescribed by the user (with probabilistic assumptions based on collected data) rather than predicted by the model and that the current models areonlysimulatingseparatedbehavioralfacts.Thisshowsthereisneedforanewmethodtovalidateexistinghumanbehaviormodelsandforanalyzinghumanbehaviorinextremebuildingevacuationsituations.

Theaimofthepresentedresearchisthustoachieveabetterunderstandingofwhatactuallyhappensduring an extreme situation and how people come to decisions within a serious gaming approach. Theresearchhypothesistobeexaminedis:Canhumanevacuationbehaviorbeexploredusinga

computergame?Whileplayingagame„real“peopleratherthansoftwareagentswillhavetoslipintotheroleoftheevacuees.Toimplementthisapproach,thechallengeistodeveloparealisticandvalidseriousgameforanewkindofimmersive,dynamicandinteractivesimulationofbuildingdisasters.Toaddressthepresentedhypothesis,scientistsattheInstituteofNumericalMethodsandInformaticsinCivilEngineering(IIB)andattheInstituteofPsychologyatTechnischeUniversitätDarmstadtareworkingtogetherinamultidisciplinaryprojectteam.Forthisresearch,theinteractionoccupant-building structure is of particular interest because this cannot really be investigated by real world experimentation.Butcreatingrealisticgamescenariosisstilltime-consuming,becauseofthelargeamountofmodelingworkthathastobeperformedtwice(e.g.,3D-modeling):firstbythearchitectfortheparametricbuildingmodelaspartoftraditionalbuildingplanning,andasecondtimebythegameartistgeneratingthe3D-gamecontent.

Thechallengesare,firstly,tomodelthegamescenariobasedonparametricbuildingobjects(geometry,structureandfurthertechnicalsemanticoutofrealbuildingdesigns),and,secondly,toenhancethemodelwithanauthenticsimulationoftheemergencyscenario(e.g.,fire,smoke,explosions).Tomakesurethatthemodelisvalid,itisessentialthatthesimulationsarecomparabletostate-of-the-artfiresafetyengineeringsimulations.Atthispointdomain-specificknowledgeespeciallyfromthefieldofcivilandfiresafetyengineeringisrequired.OnepossibilityistoretrievethisknowledgefromtheBuildingInformationModeling(BIM)processbyusingitasa„knowledgerepository“.ResearchersatIIBaredevelopingnewmethodsforbringingBIMandseriousgamingtogetherforanautomaticgenerationofthegamescenariooutoftheBIM.Anotherchallengeistostimulateanimmersiveexperienceinthegamer.Thisisimportanttoimprovethepresenceofthegamerinsidethecomputergame.Generally,virtualenvironmentsvarygreatlyinthequalityofrepresentingtherealworld.Itisplausibletoassumethatthemoreaccurateandrichlydetailedthereal-worldismappedontothevirtual-world,andthemorethesensescanbeadequatelystimulated,themoretheimmersiveeffectofavirtualenvironmentwillincrease.Totesttheseassumptions,scientistsformtheInstituteofPsychologyarecurrentlyconductinginitialexperimentsregardingthestereoimagecapabilities,soundeffectsandhumaninterfacedevices(HID)inthecontextofthisprojecttofindoptimalhardwaresettings.Thesensestobeconsideredinprinciplearehearing,touch,smell,taste,andsight.Inordertominimizeunrealisticbehaviorinthevirtualworld,thegamershouldhavethefeelingtobephysicallyinsidethevirtualworld.Playingcomputer-games,gamersoftentendtomakeunrealisticdecisionsduetolackofphysicalpainandinjuryresponsesfromthevirtualworld.Thisisknowninmilitarytrainingasthe„super-soldiersyndrome“[6,7].Thereforestrategieshavetobedevelopedtoavoidthiseffectandtoensurethatgamersdonotconsiderthemselvestobeinvulnerable„super-evacuees“.

ThepaperintroducestheconceptofanewkindofaBIM-basedseriousgame,foraninteractiveandreal-timesimulationofemergencyscenarios.Thishasthepotentialtoprovideanewmethodfordatacollection and validation in the area of evacuation simulations.

REFERENCES

[1] ISO TC 92 (2002). Res.244 TG N6 Framework for the long term standardization of fire safety in support of performance-based design.

[2] Friedman, Raymond, (1992) An International Survey of Computer Models for Fire and Smoke SFPE Journal of Fire Protection Engineering, 4 (3), 1992, p. 81-92 Available online: http://www.firemodelsurvey.com last accessed May 2011

[3] Santos G. Aguirre, B.E. (2004). A Critical Review of Emergency Evacuation Simulation Models. In NIST Workshop on Building Occupant Movement during Fire Emergencies

[4] Tavares, R.M. (2008). Evacuation Processes Versus Evacuation Models: „Quo Vadimus“?, Fire Technology 2008, ISSN-1572-8099

[5] Kuligowski, E.D.; Gwynne, S., (2010) The Need for Behavioral Theory in Evacuation Modeling, in Pedestrian and Evacuation Dynamics 2008 Isbn: 978-3-642-04504-2, Springer Berlin Heidelberg

[6] Barlow M. And Morrison, P., (2005). Challenging the Super Soldier Syndrome in 1st Person Simulations, Available online: www.siaa.asn.au/get/2411856229.pdf, Last accessed: May 2011

[7] Morrison, P., Barlow, M., Bethel, G. And Clothier, S., (2005). Proficient soldier to skilled gamer: training for COTS Available online: http://seal.tst.adfa.edu.au/research/vesl/Papers/Training4COTSSuccess_simtect05.pdf Last accessed: May 2011

Validation and calibration of the EXIT89 evacuation model for road tunnel evacuation applications

Enrico Ronchi, Politecnico di Bari, Bari ITALYRita F. Fahy, National Fire Protection Association (NFPA), Quincy MA USAPasquale Colonna, Politecnico di Bari, Bari ITALYNicola Berloco, Politecnico di Bari, Bari ITALY

ComputationalModellingsoftwarehasbeenusedinrecentyearsasatoolforanalysingtheoccupants’safetyconditionincaseofemergencyinevermorecomplexinfrastructures.Thisisthereasonwhytheirapplication-initiallyalmostexclusivelyforbuildings-iscurrentlyextendedtoalargenumberofenvironments,includingundergroundspaces.Roadtunnelsareuniqueenvironmentswiththeirownspecificcharacteristics:unknowntousers,nonaturallight,etc.,whichaffectdifferentaspectsofHumanBehavioursuchaspre-evacuationtimes(e.g.peoplemayshowvehicleattachment),occupant-occupantandoccupant-fireinteractions,herdingbehaviours,exitselection,etc.[Boer,2003andShieldsandBoyce,2004].

Thevalidationofanevacuationmodelisgenerallyperformedthroughtestingitspredictivecapabilitieswithinasetofstandardenvironments(e.g.buildings)orstandardlayoutssuchastheIMOtests(IMO,2007).Unfortunately,non-expertuserscouldconsidermodelresultsasreliableinuniqueenvironmentsaswell,andextendtheirusetoapplicationswherenoadhocvalidationtestshavebeenperformed.Theuseofamodelbeyonditsvalidationevidencerequiresthenanadditionaleffortbytheevacuationmodellertounderstandthemodellimitationsinrepresentingtheevacuationprocessinthatspecificenvironment.

ThispaperfocusesontheapplicabilityoftheEXIT89evacuationmodelforroadtunnelevacuations.EXIT89isanetwork-basedmodelabletosimulatetheevacuationoflargepopulationsthroughcomplexinfrastructureslikehigh-risebuildings.Themodelpermitsthesimulationofdifferentaspectsrelatedtohumanperformanceunderfireconditions,suchastravelpaths,delaytimes,mergingflows,counter-flows,etc.Travelspeedsareconsideredwithinthemodelasafunctionofthechangingcrowdednessofspacesduringtheevacuation[Fahy,1996and2001,ISOdocument2011].Walkingspeedsarethencalculatedinaccordancewithpeopledensityaswellasoccupantcharacteristics,usingtheequationsprovidedbyPredtechenskiiandMilinskii[1978].Roadtunnelsareenvironmentswherelowoccupantloadmayoftenoccurduringevacuations,exceptformixed-usedtunnels(i.e.pedestriansandvehicles)wherehighdensitiescaneasilyarise[Ronchietal.,2010].Thecrucialfactorsduringanevacuationscenariocouldthenberelatedtotheoccupants’behaviourduringthepre-evacuationphasemorethanonthepeoplemovementitself[Purser,2009].Thus,thereisaneedoftestingthepredictivecapabilitiesofmodelswherespeedsarebasedonpeopledensity(e.g.EXIT89)forthespecificcaseofroad tunnel evacuation scenarios.

Thefirstpartofthepaperisavalidationexerciseforroadtunnelapplications.ThepredictivecapabilitiesofEXIT89aretestedbycomparingthemodelresultswiththetunnelevacuationexperimentperformedbytheDepartmentofFireSafetyEngineeringandSystemSafetyofLund

Universityin2006[Nilssonetal.,2009]intheGötatunnelinGöteborg.Participantswerepartiallyinformedduringtheexperimentsi.e.theywereonlytoldtheyweretakingpartinastudyaboutdrivingbehaviour,thuspermittingtheanalysisofparticipants’behavioursincaseofanunannouncedevacuation.Resultsshowthatthemodelisabletoreproducetheevacuationprocess,althoughahighdegreeofmodeller’sexpertiseisrequiredforaccuratelycalibratingthemodelinputincaseoflowoccupantload.Therepresentationofthetunnelenvironmentisdonebysimulatingdifferentpartsofthetunnelthroughanetworkofnodes.Thusmodellersneedtocalibratethecorrectdimensionsofnodestotakeintoaccountthefactthemodelisnotreproducingseparatedroomsasincaseofbuildingenvironments.

ThesecondpartofthepaperisacasestudyofanItalianroadtunnel,theCondòtunnelinLecce.Itisatwo-boreroadtunnelwithanemergencylinktunnelbetweenthetwobores.ItbelongstotheTrans-EuropeanNetworkthatinItalyisunderthemanagementoftheItalianCompanyofRoads(ANASAziendaNazionaleAutonomadelleStrade).ThepredictivecapabilitiesofEXIT89wereusedtoanalysedifferentscenariosinwhichasetofdifferentvariablesaffectingtheevacuationprocesswerevariedi.e.visibilityconditions,initialwalkingspeeds,occupantload,etc.

Inparticular,thepurposeofthiscasestudyistoevaluatehowdifferentstartingvisibilityconditionswouldaffectthefinalevacuationtimes.Thisisdonebyassumingdifferentextinctioncoefficientswithinthetunnelrepresentingdifferentdegreesofseverityoftheoccurredaccidents.InitialwalkingspeedsarethensimulatedbyapplyingthecorrelationbetweenextinctioncoefficientsandwalkingspeedsprovidedbyJin[1970].Resultsprovideinformationontheevacuationprocess(e.g.mergingflows,occupantdensitiesinthenodesduringthepassageoftime,etc.)andsubsequentevacuationtimesinrelationtotheassumedvisibilityconditions.EXIT89resultsarecomparedwiththeresultsproducedbythecapacitymethodprovidedintheSocietyofFireProtectionEngineeringHandbook[GwynneandRosenbaum,2008].Particularattentionispaidtothedifferencesamongthetwomethodsemployedforscenarios considering a low occupant load.

Conclusionsfocusonmodelstrengthsandlimitationsinthereproductionofhumanbehaviouraspectsrelatedtoroadtunnelevacuations.Considerationstoimprovethereliabilityofthemodelresultsareprovided as well.

REFERENCES

[1] IMO (2007), Guidelines for Evacuation Analyses for New and Existing Passenger Ships, MSC/Circ.1238, International Maritime Organization, London, UK.

[2] Fahy, R. F. (1996) Enhancement of EXIT89 and analysis of World Trade Center data, NIST-GCR-95-684, Fire Analysis and Research Division.

[3] Fahy, R. F. (2001), Update on the Features and Demonstrated Predictive Capability of EXIT89. Engineered Fire Protection Design. Applying Fire Science to Fire Protection Problems, International Conference. Proceedings. Co-organized by: Society of Fire Protection Engieners (SFPE) and National Institute of Standards and Technology (NIST). June 11-15, 2001, San Francisco, CA, pp. 303-314.

[4] ISO/TC 92/SC 4 N622 (2011) Fire Safety Engineering – Example on verification and validation of a calculation method – Part 4: Egress model.

[5] Predtetschenski W.M., and Milinskii, A. L., (1978) Planning for foot traffic flow in buildings, Amerind Publishing, New Dehli.

[6] Boer L. C., (2003) Behavior of Drivers during Tunnel Evacuation. (Re)Claiming the Underground Space (pp 213-217) J. Saveur, Lisse, The Netherlands.

[7] Shields T. J., Boyce K., (2004) Towards Developing an Understanding of Human Behaviour in Fire in Tunnels. Proceedings of 3rd International Symposium Human Behavior in Fire, Belfast, UK pp215-228.

[8] Purser, D. A. (2009) Application of human behaviour and toxic hazard analysis to the validation of CFD modelling for the Mont Blanc Tunnel fire incident. Proceedings of the Advanced Research Workshop: Fire Protection and Life Safety in Buildings and Transport systems University of Cantabria, Spain. 15-17 October 2009.pp 23-57.

[9] Ronchi, E., Alvear, D. Berloco, N., Capote, J. Colonna, P., Cuesta, A.: (2010) Human Behaviour in Road Tunnel Fires: Comparison between Egress Models (FDS+Evac, STEPS, Pathfinder), Proceedings of INTERFLAM2010, pp. 837-848 Nottingham, UK.

[10] Nilsson, D., Johansson, M., & Frantzich, H. (2009) Evacuation experiment in a road tunnel: A study of human behaviour and technical installations. Fire Safety Journal, 44(4), 458-468.

[11] Jin, T. (1970) Visibility through Fire Smoke, Bull. of Japanese Assoc. of Fire Science & Eng., 19, 2, pp. 1–8.

[12] Gwynne, S. M. V., Rosenbaum, E. (2008) Employing the Hydraulic Model in Assessing Emergency Movement. SFPE Handbook of Fire Protection Engineering, 4th Edition. National Fire Protection Association, Quincy, MA.

Poster exhibition

Unsorted list

Li-yun Dong, Shanghai University, Shanghai CHINA Simulation of bidirectional pedestrian flow through a channel with a bottleneck

Alessandro Corbetta, Politecnico di Torino, Torino ITALY Multi-scale non-local first-order modelling of crowd dynamics: a general framework with application to footbridges

Majed Almejmaj, Worcester Polytechnic Institute, Worcester MA USA Observations from student exercises and proposed experimental scenarios for human hehavior data collection

Minjie Chen, Technische Universität Berlin, Berlin GERMANY Simulation of pedestrian dynamics with density control on a regular grid

Kongjin Zhu, University of Science and Technology of China, Hefei CHINA The non-symmetrical choice behavior during evacuation experiments in building with multi-obstacle

Georg Walenciak, Universität Heidelberg, Heidelberg GERMANY Agent-based modelling and evacuation simulation for disaster preparedness and management

Vaisagh Viswanathan, Nanyang Technological University, Singapore SINGAPORE An information-based model of pre-evacuation behavior for agent based egress simulation Masatoshi Kaitsuji, Kobe University, Kobe JAPAN Venue suitability for large-scale events from the standpoint of crowd safety

Simulation of bidirectional pedestrian flow through a channel with a bottleneck

Li-yun Dong, Shanghai University, Shanghai CHINAXiang Li, Shanghai University, Shanghai CHINAPeng Zhang, Shanghai University, Shanghai CHINA

Bidirectionalpedestrianflowsthroughachannelhavebeeninvestigatednumericallybymanyresearchersinrecentyears.Severalexperimentsontheunidirectionalflowspassingabottleneckhave been performed and some experimental data are available to calibrate these simulation models forpedestrians..Tajimaetal.investigatedtheunidirectionalpedestrianflowthroughachannelwithabottleneckbythelatticegasmodel.However,similarproblemsforpedestriancounterflowhavebeenseldominvestigatedeitherexperimentallyornumerically,especiallybydiscretemicroscopicmodels,e.g.latticegasmodelandcellularautomatamodel.Duetotheexistenceofthebottleneck,itiseasytoformcongestedareaonbothsideofbottleneckevenatratherlowdensitieswithoutefficientinteractionsduringsimulations.Helbingetal.carriedoutseveralexperimentsforcounterflowinachannelwithbottlenecks,whichisforcomparisonwithunidirectionalflow,andwhichexclusivelyindicatedmanyself-organizedphenomena,suchaslaneformation.Thesephenomenawerealsoreproducedbyadoptingthesocialforcemodel.Bursteddeetal.proposedacellularautomatamodeltoreproducecollectivephenomenabyintroducingfloorfieldinspiredbytrailformationmodelssuchaslaneformationandtheoscillationsofthedirectionofflowatbottleneck.However,theseresultsarereportedinaverysimplemanner.Hereafter,fewworkshavebeendonetosupportorimprovethesefindingsinthecontextofcellularautomatamodels.

Thepresentpapersimulatesthebidirectionalpedestrianflowthroughabottleneckatthemiddleofthechannelunderperiodicboundaryconditions.ThemodelusedhereistheextensionofthecellularautomatamodelforpedestriansbasedonthefloorfieldthatwasproposedbyBursteddeandSchadschneider.Thebottleneckcanbeviewedasanobstacle,thusthedistanceinthestaticfloorfieldisnotsimplytakenasthatbetweenapedestrian’scurrentpositiontothetarget.ThemethodsuggestedbyHuangisadoptedtoobtainthestaticfieldwhichcalculatesthedistanceoftheshortestpathtothetargetwithoutconsideringpedestrians.Twostaticfloorfieldsaregeneratedfortherightandleftwalkingpedestrians,respectively.Asaconsequence,apedestrianisdrivenbythestaticfloorfieldandwalktowardshistarget.Here,thepedestrianisonlypermittedtomoveinfourdirections,i.e.,right,left,upanddown.Accordingtothestaticfieldandthestatesofneighboringcells,onecancalculatethematrixoftransitionprobabilitiesofthepedestrianbythemethodsuggestedbySchadschneider.Inaddition,thevisualfieldisintroducedtorepresentthedynamicaleffectgeneratedbypedestrianswhichenlargestheinteractionregionbetweenpedestrians.Thescopeofvisualfieldisarectangleareaofsizem*2n+1,i.e.thesizeoftheright,frontandleftvisualfieldsarem*n,m*1andm*nrespectively,wheremisthelengthofthevisualfieldandnthewidthoftherightorleftvisualfield.Thesepedestriansofboththesameandoppositedirectionsinthevisualfieldofapedestrianaretakenintoaccount.Andthreetransitionprobabilitiesinthevisualfieldaremodifiedaccordingly.Thevisualfieldreducesthenumber of encounters with oppositely moving pedestrians and strengthens the tendency to follow

precedingpedestrianswiththesamedirection.Therefore,themovementofapedestrianisdeterminedbythemodifiedmatrixoftransitionprobabilities.Theparallelupdatingrulesareusedduringthesimulation.Oneshoulddealwithcollisionsamongpedestrians,i.e.,severalpedestriansmayenterthesametargetcellsimultaneously.Thentheyaresettohaveequalprobabilitytodoso.Atthebeginningofeachsimulation,pedestriansarerandomlydistributedatagivendensityinthechannel.

Numericalsimulationsareperformedtodrawthefollowingconclusions:(1)ItisfoundthatboththesaturatedflowrateandthecriticaldensityincreasewiththebottleneckwidthWd.Foragivenbottleneckwidth,theaveragevelocitydecreaseswiththedensity.Adynamicalphasetransitionoccursfromthefreeflowtothechokingflow.(2)TheeffectofthesensitivityparameterKsforstaticfloorfieldisinvestigated.ItturnsoutthatthesaturatedflowratesarenearlythesamewhenKsisgreaterthanorequaltoone.Whileonetakessmallervalueofvalue,pedestrianshavenostrongintentiontopassthebottleneckandleadtoalowersaturatedflowrate.ThecriticaldensitydecreasewiththesensitivityparameterKs.Itmeansthatcongestionatbottleneckoccursatlowerdensitiesifpedestrianshavestrongerintentiontopassthebottleneck(i.e.“faster-is-slower”).(3)Thesizeofthescopeofvisualfieldisstudied.Itisshownthatboththesaturatedflowrateandthecriticaldensityincreasewiththelengthofvisualfieldm,however,thewidthofvisualfieldhasnegligibleeffect.Ifthevisualfieldisnotconsidered,thesaturatedflowrateismuchlowerthanthatwithvisualfieldaswellasthecriticaldensity.Thevisualfieldplaysavitalroleinenhancingtheefficiencyoflocalinteractionsamongpedestrians.(4)Differentratiosoftwogroupswithoppositedirectionsisinvestigatedandfoundthatboththesaturatedflowrateandthecriticaldensityhavenotchangedsignificantlybuttheasymmetricalcaseenhancethefluxwhenthedensityislargerthanthecriticaldensity.(5)Thesnapshotofasimulationwithρ=0.1,Wd=5showstheseparationoftwogroupsofoppositedirectionspassingthebottleneckintotwoseparatelanesafterlongenoughsimulation.Butatahighdensityρ=0.2,pedestriansbuilduponbothsideofthebottleneckandformacongestedarea.However,thisareaisnotfullycompactsothatfewpedestrianscanreleasefromit.Usually,suchaprocessisintermittent,thereforeitcannotberegardedasthetypicaloscillationsatbottlenecks.Insummary,thevisualfieldisintroducedtothecellularautomatamodelbasedonthefloorfield,whichenhancetheefficiencyoflocalinteractionsandavoidunnecessarycollisionssignificantly.Therefore,theimprovedmodelcangiveabetterdescriptionofpedestriancounterflowthroughabottleneckinachannel.

REFERENCES

M. Muramatsu M, T. Irie and T. Nagatani, Jamming transition in pedestrian counter flow, Physica A 267, 487-498 (1999)

T. Kretz, A. Grünebohm and M. Schreckenberg, Experimental study of pedestrian flow through a bottleneck, Journal of Statistical Mechanics: Theory and Experiment, P10014 (2006)

Y. Tajima, K. Takimoto and T. Nagatani, Scaling of pedestrian channel flow with a bottleneck, Physica A 294, 257-268 (2001)

D. Helbing and P. Molnár, Social force model for pedestrian dynamics, Physical Review E 51(5), 4282-4286 (1995)

D. Helbing, L. Buzna, A. Johansson and T. Werner, Self-organized pedestrian crowd dynamics: Experiments, simulations, and design solutions, Transportation Science 39(1), 1-24 (2005)

C. Burstedde, K. Klauck, A. Schadschneider and J. Zittartz, Simulation of pedestrian dynamics using a two-dimensional cellular automaton, Physica A 295, 507-525 (2001)

H. J. Huang, R. Y. Guo, Static floor field and exit choice for pedestrian evacuation in rooms with internal obstacles and multiple exits, Physical Review E 78, 021131 (2008)

Observations from student exercises and proposed experimental scenarios for human hehavior data collection

Majed Almejmaj, Worcester Polytechnic Institute, Worcester MA USABrian Meacham, Worcester Polytechnic Institute, Worcester MA USA

Experimentsandfieldobservationshavebeenaprimarysourcefordatausedtodevelopempiricalmodelsandinputsforcomputersoftwareforevacuationsimulation(Fahy,2002).Althoughresearchershavebeenhighlightingtheneedtogatheradditionaldatafromvariousbuildingandoccupancytypes,acrossabroaderspectrumofoccupantdemographics,dataremainscarce(ProulxG.,2008).Therehavebeensomeattemptstodevelopageneralframeworkfordatacollection,especiallyduringfiredrills(GwynneS.M.,2010;ProulxG.,1996).However,specificsaboutthearchitecturalsettingsandoccupancytypeareoftennotaddressed.Thisisunfortunate,astheseareimportantfactorsthatcan assist in identifying certain behaviors which can be implemented in various computer evacuation models,suchasinfluencesonexitselection,occupancy-relatedindividualandgroupbehavior,andpotentiallyissuesassociatedwithdifferentculturalnorms. Asameanstohelpstudentslearnmoreabouttheimportanceofvarioushumanandbuildingfactorsonoccupantuseofegresssystems,andtohelpgenerateadditionaldataforanalysisanduseinevacuationmodeling,studentsintheWorcesterPolytechnicInstitutecourse,FPE580M,PeopleandFires,wereaskedtoundertakeanexperiment-basedprojectrelatedtohumanbehaviorandpedestrianmovement.Inthisexperiment-basedproject,studentswereaskedtoidentifyahumanbehaviorormovementattributeofinterest,observepeopleinpublicspacesundernormalconditions,collectandanalyzethedata,andpresenttheirfindings.Tohelpguidethestudents,fourpotentialscenarioswithminimumrequirementsforthespaceandnumberofoccupantsweresuggested(Thesescenarioswerenotrequiredtobeusedsincethecourseincludedon-campusanddistancelearningstudents,andflexibilitywasneededtoaccommodatedatacollectioninoffcampusenvironments).Intheend,datawerecollectedoverarangeofdifferenttypesofenvironmentssuchasuniversityclassrooms,auditoriumsandtransportationterminals.Whiletheoutcomesoftheseexercisesdonotnecessarilyconstitutecompleteorrepeatabledatasets,someinsightsweredeveloped,whichmayhelpinformfutureexperimentalplansanddatacollectionexercises.Thispaperdiscussesthesuggestedexperimentalsetups,someofthestudentprojectfindings,someofthechallengesfacedduringdatacollectionandanalysis,andsuggestionsforothersembarkingonsimilarexercisesinthefuture.Someobservationsofpotential interest include the following.

Impactofcounter-flowonmovementspeeds.Counter-flowisaconcernwhenanytypeofsimultaneousbi-directionalflowmayoccurwithinthesamecomponentsofthemeansofegresssystem.Intwoseparatestudies,studentsassessedtheeffectsofcounterflowwithinacorridoronoccupants’walkingspeed.Inbothstudies,occupantmovementspeedwas10%to50%lowerthanthedatareportedintheSFPEHandbook,evenatlowoccupantdensities.Occupantage,buildingarchitecture,anduseofthebuildingsalllikelyplayedarole.

Exitselection.Occupantstendtochoosetoevacuateaspacethroughafamiliarrouteduringanemergencysincethatitiseasierthantheprocessoffindinganotherrouteandinvolveslesseffort(Keating,1985;Passini,1984).Thoughitissuggestedthatoccupantredirectingtoanotherexitdependsonlengthofthequeueanddistancetotheexit(Gwynne,Galea,Owen,Lawrence,&Filippidis,1998-99),someofthestudents’observationsshowedotherwise.Intwostudiescoveringdifferentuniversitylecturehalls,thestudentsobservedthatthemajorityoftheoccupantschosetoevacuatethroughthefamiliarexit,anddidnotchangetheirexitchoiceevenwhentherewerequeues.Inaddition,theegressroutetakenappearedtobemorecloselyalignedtotheir‘exitofchoice,’ratherthanshortesttraveldistancetoanexit,whichiscontradictorytomanyalgorithmsusedinegressmodels.

Movementspeedonstairsandescalators.Stairsremaintheprimarymeansofegressfrommulti-storybuildings.However,muchoftheavailabledataisoutdatedandmaynotbeapplicablenowadaysduetochangesindemographics(ProulxG.,2008).Thisalsoappliestoescalators,whichareusuallynotconsideredasofficialpartsofegresssystemsduringemergencies.Instudiescoveringoccupantmovementandbehavioronstairsandescalators,studentsobservedmovementspeedsunderavarietyofconditions,suchascarryinglargeobjects,walkingingroups,andusingmobiledevices.Outcomessuggestthatcurrentdatamynotreflectactualspeeds,andillustratesareaswhereadditionresearchcouldbebeneficial.

Datacollectionforoccupantegressbehaviorscanprovidemuchneededdatatotheengineeringandresearchcommunities.Whileforsometheidealsettingfordatacollectionwouldbethroughactualfireincidents,itissuggestedthatobservationaldataforbehaviorandmovementduringnormalconditionscanstillyieldvalidinsights,notonlyforadditionalresearch,butasinputformodelvalidationandengineeringassessment.Tofacilitatethecollectionofcommontypesofdata,whichcanbereadilycompared,assessedforgoodness,andusedinresearchandengineeringanalysis,itwouldbehelpfultoidentifyacollectionofcommonexperimental/observationalsetupsforusebythe research and academic communities. It is further suggested that the common experimental setups couldbeemployedatavarietyofuniversities,providingthepotentialforcomparativedataanalysisacrossdifferentsocietiesandcultures.Whiletheoutcomesoftheabovementionedexercisesdonotnecessarilyconstitutecompleteorrepeatabledatasets,somevaluableinsightsweredeveloped,anditis hoped that lessons learned and insights gained will help inform future experimental plans and data collection exercises.

REFERENCES

Fahy, R. F. (2002). Available Data and input into Models. National Research Council Workshop to Identify Innovative Research Needs to Foster Improved Fire Safety in the United States, Session on Human Behavior. National Research Council.

Gwynne, S. M. (2010). Conventions in the Collection and Use of Human Performance Data. NIST GRC 10-928.

Gwynne, S., Galea, E. R., Owen, M., Lawrence, P. J., & Filippidis, L. (1998-99). Daptive Decision-Making in Response to Crowd Formation in buildingExodus. Journal of Applied Fire Science, 301-325.

Keating, J. P. (1985). Human Response During Fire Situations: A Role for Social Engineering. Research and Design 85: Architectrual Applications of Design and Technology Research (pp. 285-288). Los Angeles: American Institute of Architects.

Passini, R. (1984). Wayfinding in Architecture. New York: Van Nostrand Reinhold Company Inc.

Proulx, G. (1996). Methodology for Evacuation Drill Studies. Ottawa, Ontario: National Research Council Canada.

Proulx, G. (2008). Evacuation Time. In S. o. Engineers, The SFPE Handbook of Fire Protection Engineers (pp. 3-355 to 3-370). Bethesda, MD: National Fire Protection Association.

Simulation of pedestrian dynamics with density control on a regular grid

Minjie Chen, Technische Universität Berlin, Berlin GERMANYGünter Bärwolff, Technische Universität Berlin, Berlin GERMANYHartmut Schwandt, Technische Universität Berlin, Berlin GERMANY

Discretemodellingofpedestriandynamicsoftendefinesthesystemgeometryonatwo-dimensionalregulargrid.Inthesimulation,wemustconsiderthattheindividualpedestriansareassociatedwithanexclusivepersonalspace.Inthetraditionaltwo-dimensionalcellularautomatonmodel(CA)anditsvariousextensions(see[1],[4],[10],[5],[7],[6],[8]etc.)thispersonalspaceisdescribedbythecellsizeoftheunderlyingregulargrid.Hence,thestatechangeonthegridcanbeappliedtodescribethesystemdynamics.Thisleadsinevitablytoafixedpersonalspaceofthepedestrians.However,empiricaldata(e.g.[9])showthatinlowdensityrange,thesizeofthispersonalspacevariessignificantly.Inhighdensityrange,thesizeisclearlyrestrictedbythephysicalsizeofthepedestrians.Thepurposeofthispaperistopresentanewmodellingtechniqueofpedestriandynamicswithconsiderationofadvancedstepcalculationanddensitycontrolonatwo-dimensionalregulargrid.

Themaincontributionofthepaperwillbeexplainedinthefollowingparts.

1.In[3]weproposedanewmethodforthestepcalculationonaregulargridforthegeneralizedcaseoflocalvelocitylargerthanonegridcellsizeinonedimensionpersimulationcycle(sometimesalsocalledmulticell-step),whichisnecessitatedbytherepresentationofheterogeneouspedestrians.Thelocal step choice on the underlying grid is more than a simple position change from a start position toadestinationcalculatedbya“hardware”method,e.g.oftheshortestdistance,withouttakingintoconsiderationtheactualsystemdynamics.Instead,weallowsmalldeviationsonroutewithinalocalstep,butwiththerestrictionthatthemathematicalexpectationofthepossiblestepchoicesshouldreflect the original position transition exactly.

2.Weintroduceaprojectionmechanismtocomputetheintermediatestepsforthepositiontransitionwithconsiderationoflocal(cell)positionavailabilities.Inthesimulationcycle,wefurtherintroduceabalancingmechanismtodecidetheexecutionsequenceofthesimulationparticipants.Thisimprovement,alongwiththefirst,enablesadrasticreductionofthepossible“deadlock”amongtheparticipants and at the same time gives a very reasonable explanation for the step calculation.

3.Inreality,however,pedestrians’personalspacesometimesappearstobe“compressible”toacertainextentinawaysimilartogasparticles.Toourknowledge,intheCAmodelanditsextensions,duetothenatureofhomogeneousrectangles(orsquares)onthetwo-dimensionalregulargrid,thepedestrians’exclusivepersonalspaceisfixedinthesystemconstruction,e.g.as0.4m*0.4min[1].Weproposedin[2]asolutionforaflexiblepersonalspace.Weobservethatthisexclusivepersonalspacevariesroughlyintherangeof0.3m*0.3m(whichrepresentsthe“incompressible”physicalsizeofapedestrian)to1m*1m(whichcorrespondstothenormal,unstressedstateofanaveragepedestrian)

inreality.Henceinthecurrentpaper,thegridcelloftheregulargridisdefinedtobeofsize0.3m*0.3m,andthemodifiablepersonalspaceforthepedestriancanbeconsideredasacompositionofexactlyonegridcelltoaneighbourhoodof3*3gridcellsontheregulargrid.Weconsidertwocases.

(a)Firstly,thespacerequirementofasinglepedestrianvariesempiricallyincorrelationwithlocalspeed.Thatis,ahigheroverallspeedinthesystemresultsinalowerdensity.Thisissometimesreferredtoasfundamentaldiagram.Thefundamentaldiagramenablesustoestimatetheempiricaldensityviathelocalpedestrianvelocity,whichcanbeacquiredfromthelastsimulationcycle.Accordingtothisinformation,someoftheeightgridcells(apartfromtheorigin)inthecorresponding3*3-neighbourhoodaretobedeclaredas“inaccessible”fortherestofthesimulationparticipants.Thisinaccessibility will be described by a transition function.

(b)Secondly,theformercaseconsidersthesituationthatapedestrianbehavesonhisorherownbehalf.Inanextensionofthis,wedistinguishthenotionof“inaccessibility”undervariouscontexts:Independentpedestrianstendtoshowgreater“repulsive”effecttoeachother,whereaspedestrianswhichbelongtothesamegroupinthesimulationenvironmenttendtobe“friendly”toeachother,hence,theinaccessibilityfunctionisofalowervalue.

Overall,theinaccessibilityatanarbitrarycellpositionisalwaysdependentonthebehaviouralcontextinthesystem.Thedensitycontrolonalocallevelcanbeachievedinthisway.

Wepresenttwoexperimentalcasestodemonstrateourmodel.Inthefirstexperimentwesimulatethegroupbehavioursofthepedestrians,i.e.howsomeofthesetendtostayclosetoeachother(butthedistanceamongthemmaystillvaryoverasmallrangeinthelocalstepcalculation),distinguishingthem from other “foreign” pedestrians which exist in the simulation environment at the same time. Asecondexperimentconsiderstheintersectionofmultiplepedestriansgroupswithasimplificationthatalltheparticipantsaregivenpre-configuredmovingdirections.Theobjectiveofthisexperimentistodemonstratethedensityevolution,especiallyincertaincriticalregions(e.g.theintersectionpointwherebottleneckphenomenomismorelikelytotakeplace),incorrelationwithlocalspeed.

REFERENCES

[1] C. Burstedde, K. Klauck, A. Schadschneider, and J. Zittartz. Simulation of pedestrian dynamics using a two-dimensional cellular automaton. Physica A, 295:507–525, 2001.

[2] M.-J. Chen, G. Bärwolff, and H. Schwandt. Automaton model with variable cell size for the simulation of pedestrian flow, 2008. An electronic version can be retrieved at: http://page.math.tu-berlin.de/~chenmin/pub/cbs080331.pdf (accessed November 10, 2011).

[3] M.-J. Chen, G. Bärwolff, and H. Schwandt. A study of step calculations in traffic cellular automaton models. In 13th International IEEE Conference on Intelligent Transportation Systems, pages 747–752, 2010. An electronic version can be retrieved at: http://page.math.tu-berlin.de/~chenmin/pub/cbs100709.pdf (accessed November 10, 2011).

[4] A. Keßel, H. Klüpfel, J. Wahle, and M. Schreckenberg. Microscopic simulation of pedestrian crowd motion, pages 193–200.

[5] A. Kirchner, K. Nishinari, and A. Schadschneider. Friction effects and clogging in a cellular automaton model for pedestrian dynamics. Physical Review E, 67(5):056122, 2003.

[6] A. Kirchner, H. Klüpfel, K. Nishinari, A. Schadschneider, and M. Schreckenberg. Discretization effects and the influence of walking speed in cellular automata models for pedestrians dynamics. Journal of Statistical Mechanics: Theory and Experiment, 2004(10):P10011, 2004.

[7] H. L. Klüpfel. A cellular automaton model for crowd movement and egress simulation. PhD thesis, Universität Duisburg-Essen, 2003.

[8] K. Nishinari, A. Kirchner, A. Namazi, and A. Schadschneider. Extended floor field CA model for evacuation dynamics. IEICE Trans. Inf. and Syst., E87-D(3):726–732, 2004.

[9] M. Plaue, M.-J. Chen, G. Bärwolff, and H. Schwandt. Trajectory extraction and density analysis of intersecting pedestrian flows from video recordings, 2011. Proc. PIA 11, LNCS 6952, 285–296, 2011.

[10] A. Schadschneider. Cellular automaton approach to pedestrian dynamics-theory, pages 75–85.

The non-symmetrical choice behavior during evacuation experiments in building with multi-obstacle

Kongjin Zhu, University of Science and Technology of China, Hefei CHINALizhong Yang, University of Science and Technology of China, Hefei CHINAXin Zhan, University of Science and Technology of China, Hefei CHINAPing Rao, University of Science and Technology of China, Hefei CHINA

Sincethefatalitiesaccidentsinbuildingfiresoccurmorefrequentlyandthesafetyissuesofthelager-scalecrowdsthatgatherinaplaceforentertainment,ceremoniesoreducationbecomesanimportantelement,moreandmoreattentionhasbeenpaidtounderstandhumanbehaviorinemergencyandoccupantevacuationinbuildings,furthermore,considerableevacuationmodelsandsoftwarehavebeendeveloped.Thesemodelsaremainlyusedtoreproduceoccupantevacuationprocessandpredictevacuationresults(suchasevacuationtimeandoccupantflowrate)inspecificconditions.However,thecalibrationandvalidationofthemodelsneedlotsofconvincingempiricaldata.Unfortunately,the reliable data and information with regard to pedestrian and evacuation dynamics is still scarce. Additionally,therearefewmodelsfocusingonreproducingandexploringthespecifichumanbehaviorduringevacuationprocess,suchasselectionbehavior.

Inthispaper,wewillstudycarefullythehumanchoicebehaviorincludingexitselectionandaisleselectionbasedonthecontrolledevacuationexperiments.Wehaveconductedaseriesevacuationexperiments.Therewere102collegestudentstookpartintheexperimentsasevacuees.Theplacecarried out the experiments was a university building and the experimental area consisted of a classroom,apassageandtwostairs.Duringtheexperiments,ateachendofthepassage,therewasanalarmwhichcangiveaudibleinformationrepresentingstairexitstatus(openorclosed).Whatneedstobeemphasizedisthattheinformationmaybefalsealarmornotactivatedwhichwecontrolledintentionallyasavariabletostudytherespondofpeopletoaudileinformationduringevacuations.Theexperimentvariablesconsistedofalarmstatus,statusofthestairexit,andthetypeofoccupantinitialdistribution.Thereweretwotypesofoccupantinitialdistribution.Oneisnonsymmetricaldistribution,thatis,thereweremoreindividualsinthebackoftheclassroomthanthatinthefrontoftheclassroom.Andtheotherissymmetricaldistribution,namely,allthestudentsdistributedintheclassroomnearlyevenly.Wepositionedtotallyninecamerastorecordtheexperimentprocess.

Bycontrollingtheexperimentalvariables,weconductedtotallyeightexperimentswhichweredividedintotwocategories:withoutandwithparticulargroups.Forwithoutparticulargroupssituations,allindividualshadthesameinstructions,atthebeginningoftheexperiment,allstudentssatontheirownpositionintheclassroom,andstartedtoevacuateassoonaspossiblewhentheaudilestartsignalwasgiven.Foreachevacuee,theevacuationfellintothreesuccessivephases.Firstly,he/shehadtomoveoutoftheclassroom,andthenchoseamovementdirectioninthepassage,moveddowntotheendof the passage and then went downstairs to the destination which was under the next below floor. If someonemovedtoaclosedstairexit,he/shehadtomovebackandevacuatedviatheotherstairexittothedestination.Allparticipantsworeredhatinordertoindentifyclearly.Forwithparticulargroupssituations,wechoserandomlyseveralstudentsasparticularindividuals,andtheyweretoldthatthey

hadtomeettogetherattheareaneartheclassroomexitinthepassagebeforetheymovedtostairexit;furthermore,theyhadtokeeptogetherduringthewholeevacuationprocess.Theordinaryindividualswerethesameasthoseinthefirstcategory,whowerejusttoldtoevacuatetodestinationasfastaspossible.Theordinaryindividualsworeredhats,andtheparticularindividualshaddifferentcolorhatsaccordingtodifferentgroup.Inaddition,inthesecondcategoryofexperiments,therewasonlyonestairexit(stairB)available,anditwastoldtoallbeforeevacuation,sotheyneednottoestimatewhichescapedirectionwasavailablewhentheyarrivedatthepassage.Thestudentswereveryfamiliarwiththebuilding,andhighlymotivated.Duringtheexperimentprocess,theywereaskedmovingasfastaspossible.Aftereachexperiment,allstudentscamebacktotheclassroom,andtherewasfiveminutestorest,sotheydidnotrepresentfatigueduringtheexperiments.Ifanyonefelttired,he/shecandropoutbeforeeachexperiment.Ateachexperiment,eachindividualsitondifferentinitialposition.

Basedonthecontrolledexperiments,wefindthatitisnon-symmetricalforpedestrians’exitselectionandaisleselectioninclassroom.Inthefirstcategoryofexperiments,eventhoughthereweremoreindividualssituatedinthebackoftheclassroom,thechoicepercentageofclassroomfrontexitAwaslargerthanthatofclassroombackexitB.Thisdeviationwasmoreapparentforthesecondcategoryofexperiments.Differently,thereweremuchmoreindividualschosetheclassroomexitB.Amassivenumberofpedestrianschosethemiddleaisleswhichwerebetweenseats,however,onlyfewpeoplechosetheaisleswhichwerenexttothewall.Inaddition,bifurcationpoint,wherepeoplelocatedatthesamerowdividedintooppositedirection,wasbiasedtowardstheclassroomexits.

Wefoundthatpedestriandestinations,thatis,thechoiceofstairexithadsignificantlyeffectsonoccupants’choiceofclassroomexitwhichwaspedestrians’intermediatedestination,buttheinitialoccupantdistributionhadlittleinfluence;pedestrianslocatedinmiddleareatendedtochoosetheaislewhichisneartheexitatthecostoftakingmorelateralmovementbetweenseats.

Toexplorethemechanismofthenon-symmetricalchoicebehavior,wewillsimulatetheevacuationprocessbasedonourproposedcellularautomatamodel[1-2]andamicroscopicsimulationtoolcalledNOMAD[3]whichwasdevelopedintheTransport&PlanningdepartmentoftheDelftUniversityofTechnology.Inthestudy,wewillreproducethenon-symmetricalchoicebehavior,andthenwewillcompare the results between experiments and simulations by considering the factors we discussed above(initialoccupantdistribution,expecteddestination,psychologybehavior)tofindwhichfactorsis most important.

Theresultsinthispaperareexpectedtoprovidevaluableadviceforlarge-scalecrowdmanagementbothinnormalandemergencysituationsandtheoptimizationdesignofinfrastructurelayoutinmulti-obstaclesbuildings(suchasclassroom,theatre,andstadium).

REFERENCES

[1] Yang, L.Z., et al., Cellular automata pedestrian movement model considering human behavior. Chinese Science Bulletin, 2003. 48(16): p. 1695-1699.

[2] Yang, L.Z., et al., Simulation of the kin behavior in building occupant evacuation based on Cellular Automaton. Building and Environment, 2005. 40(3): p. 411-415.

[3] http://citg.tudelft.nl/index.php?id=18769andL=1.

Agent-based modelling and evacuation simulation for disaster preparedness and management

Georg Walenciak, Universität Heidelberg, Heidelberg GERMANYGregor Lämmel, Technische Universität Berlin, Berlin GERMANYMichael Zielke, Technische Universität Berlin, Berlin GERMANYHubert Klüpfel, TraffGo HT GmbH, Duisburg GERMANYPatrick Gessler, TraffGo HT GmbH, Duisburg GERMANYAlexander Zipf, Universität Heidelberg, Heidelberg GERMANY

INTRODUCTIONAsperiodicwildfiresinthesouthernCalifornia,USA,orthenuclearFukushimaDaiicinucleardisasterhaveshown,man-madeandnaturaldisastersmightmakehumanareasorregions(temporarily)inhabitableorthesafetyofinhabitantsmightbethreatenedseverely,andtheevacuationofthisarearequired.

Onemajordecisioninfacingdisastersiswhethertoevacuateornot.Criteriaformakingthatdecisiononacivildefenselevelare:thetimeavailableforevacuation,thesocio-economicsituation,thewarningsystemsavailable,theseverityandtimeevolutionofthethread,andsoon,andsoforth.Onemajorquestioniswhethertheavailablesafeevacuationtime(ASET)islargerthantherequiredsafeevacuationtime(RSET).Anotheroneisthefeasibilitytrafficpatternsthatwillresultfromanevacuation warning.

Inthiscontext,theGRIPS-Project(GIS-Based-Risk-analysis-InformationandPlanningSystem)(http://www.geog.uni-heidelberg.de/forschung/gis_grips.html)aimsatprovidingusefulandobjectiveinformationforthedecisionmakersfacingsuchquestions.Therefore,themulti-agenttransportsimulationtoolkitMATSim(http://www.matsim.org/)isusedtosimulatetheevacuationregardingtwouse-cases.ThesimulationoftheevacuationofthedistrictWilhelmsburginHamburg,Germany,inconsiderationofanupcomingflooding.Theseconduse-caseexaminesthesimulationoftheevacuationofanareaconsideringanindustrialaccidentinthecityofEssen,Germany.Inneithercase,thedisasterismodeleddirectly,buthandledasanexternalparameter“threat”thataffectsacertainregion.Oneofthemaingoalsoftheprojectisthedevelopmentofacomputertoolthatcanbeusedbydisastermanagersasadecisionsupportsystem.Forthatmatter,bothuses-casesarespecifiedincooperationwithlocalauthorities.Generallyspoken,twocentralaspectsareofinterest:ontheonesidethetoolmightbeusedshortlybeforeadisasteroccurred.Then,civildefenseauthoritieshavetodecidewhetheradistrictshouldbeevacuatedornot.Ontheothersidethetoolmightbeusedinpre-disasterplanning,i.e.adaptingthetrafficandcivildefenseinfrastructure(likeshelters).

Fromthescientificperspective,therearetwocentralissues:extendingtheexistingagentbasedqueuemodelfortransportsimulationbyareal2Dsimulationforsimulatingpedestriansonthesmallscaleand the development of a generic data model which describes a wide variety of evacuation simulations.

AGENT-BASED-SIMULATIONExistingmodelsareeitherfocusingonthesimulationoflargeareasbutsimplifyinggeometricdetailsortheyarefocusingonsmall-scalescenarioswithhighgeometricresolutionsbutarenotcapableto

simulatelargerareasbecauseofthecomputationalcomplexity.Inthisproject,wearefocusingonamultimodalsimulation,whereeachmodeoftransportcanbesimulatedinadifferentphysicalmodel.Forvehiculartrafficitseemstobereasonabletouseasimulationmodelwithacoarseresolutionasforexamplethewellknownqueuemodel.However,formorecomplexsituationsthequeuemodelistoocoarse.Thisisusuallythecasewhenitcomestopedestrianswhoarenavigatingthroughcomplexenvironmentsliketrainstations,wherehundredsoreventhousandsofpedestrianstrytogetonacityrailway.ThisisaparticularprobleminHamburg,sincethecityrailwayispartoftheauthoritiesevacuation plan.

Whendoingmultimodalsimulationsthereisalsoaneedtohavemodechangeoptions.Forexampleapedestrianmightstartathomeandwalktothenextbusstop.Aftershehasreachedthebusstop,shehastowaitforthebus.Whenthebusarrivesshehastogotonthebus.Nowletsassumethebustakeshertothenexttrainstation,wheresheleavesthebusandwalkstotheplatformtogetonacityrailway.Thecityrailwaythenfinallytakesheroutoftheevacuationarea.Whenmodelingmodechangesthephysicalsimulationmodelsnotonlyhavetobesynchronizedbutalsotheboundaryconditionshavetobemodelcarefully.Thisishighlynon-trivialwhencouplingmodelswithdifferentgeometriccomplexity.

DEVELOPMENTOFAGENERICDATAMODELDESCRIBINGEVACUATIONSIMULATIONThedevelopmentofthetoolincludestwobasicuse-cases,butingeneralthetoolissupposedtosatisfyawidevarietyofpossiblecases.Onegoalisthedevelopmentofadatamodelwhichissuitableforavarietyofscenarios.Forthatmatter,thescopeofthedatamodelneedstobedefined.Itneedstobedefinedwhichtypesofscenariosshouldbesatisfiedandwhichscenarioaretobeoutofscope.Oneimportantaspectregardingthescopemaybethescale.Forthat,theevacuationofaplaneorasinglebuildingisconsideredtobeoutscopebecausethetooladdressesregionalevacuationonly.AnotheraspectisbetheunderlyingdatamodelsincetheevacuationsoftwareGRIPSusesatopologicalnetworkanddoesnotdirectlyworkonrasterdata.Thoseaspects,asthesemanticofdata,dataformats,consistency,qualityetc.willbeanalyzedindetailinthefullversionofthepaper.

Besidesthescope,thenecessaryinputisanalyzed.Fornow,theminimuminformationforperforminganevacuationsimulationinGRIPSisthegroundinginfrastructure,thepopulationcharacteristicsandtheevacuationarea.Regardingthoseinformation,thedatastructuremustbespecified.Forexample,thegroundinginfrastructuremightbeseenasaroadnetworkrepresentedasdirectedgraph.Inanycase,thenetworkmustbesuitableforrouting.However,thereisdifferentadditionalinformationrequiredofpedestrianrouting,automobileroutingorpublictransportrouting.Generallyspoken,differentscenariosrequiredifferentinputinformation.Thedatamodelneedstodefinetheinputdatawhicharemandatoryandwhichdataareoptional.Furtherthemodelshoulddefinesomeimportantscenariotypesforexamplepedestrianevacuation.Inthefullversionofthepaper,therewill be represented some basic aspects that are to be regarded developing a data model for evacuation simulation.

CONCLUSIONThedevelopmentoftheGIS-basedRisk-analysisInformation-andPlanningSystemwillderivenewfindingsintheareaofagent-basedevacuationsimulation.Forthatmatter,thedevelopmentofthedata-modelwillleadtosomeconclusionsontheimportantaspectsintegratingdataintoevacuationsimulation.Furthertheworkonmultimodalsimulationmodelswillexpandthefieldofapplicationofevacuation simulations to a wide area of scenarios.

An information-based model of pre-evacuation behavior for agent based egress simulation

Vaisagh Viswanathan, Nanyang Technological University, Singapore SINGAPOREMichael Lees, Nanyang Technological University, Singapore SINGAPOREHeiko Aydt, Nanyang Technological University, Singapore SINGAPORE

Ideally,whenafirestartsafirealarmgoesoff;alloccupantshearthisalarmandusethenearestsafeexittoleavethebuilding.However,thisishardlythenorm.Inmanycases,occupantsaredesensitizedfromhearing false alarms and often do not start to evacuate until they are completely sure that it is needed. OnJanuary19,2000,afireinBolandHallinSetonHallUniversitykilledthreestudentsbecausetheyhadignoredthefirealarmsassumingtheywerefalse.Thisuncertaintyabouttheauthenticityofthefirstsignofdangerisn‘tanisolatedincident[1,2].Hence,whenstudyingthebehaviorofevacuees,itisnecessarytostudyandunderstandtheiractionsfromthetimeatwhichthefirestartedrightupuntilthepointwherethelastpersonevacuated.Modelingandsimulationisoneapproachforanalyzingandunderstanding egress behavior.

Softwarethatsimulatescrowdegressisnecessarilyverycomplexbecausecrowdegressfromabuildingisitselfaverycomplexsystemwithlotsofinteractingelements(people,fire,alarms,etc.)eachofwhichcancausedifferentcomplicationsinthesystem.OneofthemostpopularmethodsforstudyingandmodelingcomplexsystemsisthroughAgentBasedModels(ABM).InABMasetofheterogeneous,intelligent entities called agents are programmed with behavior approximating humans and placed in apartiallyobservableenvironment.Asynchronousinteractionsbetweenagentsresultinmacro-leveldynamicswhichcanhelpobserverslearnmoreaboutthesystem.Somestudies[3]haveexpressedtheirreservationsaboutexistingcomputationalmodelsofegress.Thisisbecausemanymodelsmakeassumptionsabouthumanbehaviorthatstandagainstevidenceobtainedthroughextensiveexperiments in social sciences.

Onesuchassumptionisregardingpre-evacuationbehavior.AsintheBolandhallincident,peopledon‘tstartevacuatingthebuildingassoonastheyhearanalarm;eachcuethatheobserveshasacertainimpactonthepersondependingonhisidentity,socialroleandthecircumstances.Theprocessofpre-evacuationbehaviorconsistsoftwophases[4].Thefirstphaseinvolvestheagentperceivinganeventtobeoutoftheordinary.Thesecondphasehastwoparts:Thefirstistheprocessoftheagentidentifyingthesituationasafireandthesecondpartistheprocessofquantifyingtherisktoselfandotherswhichinturndeterminestheplanofaction.In[4]Kuligowskipresentsalistofvariousfactorsandhoweachofthemaffectsaperson.Thesefactorsarecategorizedintothree:subjectivepre-eventfactors,subjectiveeventfactorsandcue-basedfactors.Theeffectofeachfactori.e.whetheritincreasesordecreasesthelikelihoodofthepersonchangingfromonephasetothenextisalsoclearlydocumented.

Ahumanisalwaysprocessinginformation.However,hecanonlyprocessalimitedamountofinformation;someofthispertainstocollisionavoidance,othersarecuesandyetothersaresimplyobservationsabouttheenvironment.Thesecuesarecognitivelyprocessedbytheagentonlyifthe

amount of information they convey is great enough to be noticed in the sea of information that the agentisexposedto.Forexample,someonewhoisreadingisunlikelytonoticesmokeformingoutsidehis window until it is strong enough (i.e. gives enough information) to be noticed.

Inthispaper,wepresentabehavioralmodelforABMofcrowdegress.Thisbehavioralmodelbuildsontheideasofpre-evacuationpresentedin[4]bymodelingevacueesasserialinformationprocessors.Theevacueesidentifyandprocessinformationintermsofeventcueswhichexistthroughouttheenvironment.Cuesthatconveyenoughinformationinfluencetheagent‘seventknowledgebasewhichismadeupofdifferentbuckets;eachcorrespondingtoaparticularphaseofevacuation.Eachbuckethasapredeterminedagent-specificthreshold.Eachcuethatisperceivedbytheagentischeckedforsignificanceanddependingonitseffect(asdescribedin[4])isthenputintotheappropriatebucket(s).Ratherthanthecueitself,thecharacteristicsofthecue(frequency,ambiguity,itssource,etc.)areimportant.Onceathresholdforaparticularphaseofbehavioriscrossed,atriggerissenttotheplannertoeffectastrategychange.

Evacueesdonothaveanindefinitememoryofcuesthattheyperceive.Duringanevacuation,anevacueeonseeingsmokeasecondtimemightgetmorestressedeitherbecausehehadforgottenaboutseeingsmokeearlierorbecausethesmokereinforceshisfeelingofdanger.Atthesametime,ifhehasjustseenalotofsmoke,seeingalittlemorewouldhavenoimpact.Thisideaismodeledbyimplementingacuememory.Acuethatisinthememoryoftheagenthasnoeffectonthebuckets.Acueresidesinthememoryoftheagentforonlyashortperiodoftime.However,ifperceivedenoughtimes,thecueiswrittenintothememorypermanently.Inthemodelweincludecommunicationoftheevacueeswherebyagentscanexchangeeventknowledgeintheformofcuesandenvironmentknowledge.Exchangedinformationhasa“trust”valueassociatedwithit,dependingonthesource,which determines whether it is added to the receiver’s memory or not.

Thepaperillustratestheimportantofpre-evacuationbehaviorandacommunicationsystemthroughexperimentation.Theseexperimentsevaluatetheimpactthatpre-evacuationbehaviorcanhaveontheoverallefficiencyofegressstrategy.Weshowthatwithpre-evacuationbehavior,informationexchangeand communication plays a critical role in the overall process and illustrate that egress strategy may be betterdesignedtofacilitateinformationexchangeaswellasefficientmovementofevacuees.

REFERENCES

[1] Proulx, G. (1995). Evacuation Time and Movement in Apartment Buildings. Fire Safety Journal, 24, 229-246.

[2] Purser, D. (2001). Quantification of behavior for engineering design standards and escape time calculations. Safety Science, 38(2), 157-182.

[3] Santos, G.; Aguirre, B. E. (2004). A Critical Review of Emergency Evacuation Simulation Models.

[4] Kuligowski, Erica D. (1999). The Process of Human Behavior in Fires. NIST Technical Note 1632.

Venue suitability for large-scale events from the standpoint of crowd safety

Masatoshi Kaitsuji, Kobe University, Kobe JAPANAkihiko Hokugo, Kobe University, Kobe JAPAN

Majorcrowddisastersthathaveoccurredsincetheturnofthe21stcenturyinclude:theaccidentatthe“32ndAkashiCitizens’SummerFestival”inJapaninJuly2001,whichcaused11deathsand248injuries;theaccidentatthe“LoveParade”inDuisburg,GermanyinJuly2010,whichkilled21peopleandinjuredover500people;andtheaccidentatthePhnomPenh“WaterFestival”inCambodiawheremorethan500peoplewerekilledandmorethan600peoplewereinjured.Analysisoftheseaccidentsrevealedvariousfactors:crowdaccidentsareverylikelytobecausedbyaninterruptedflowofhigh-densitycrowd,butsettingofcrowdtrafficcontrollinesduringthesecurityplanningstagetopreventtheinterruptionofhigh-densitycrowdflowwouldbeunrealisticbecausethetotallengthofsuchlinewouldverylong.Inaddition,evenifthebestpossiblesafetymeasuresaretakenattheeventsite,themovementofthecrowdcouldexceedthesecuritycapability,makingitdifficulttoaverttheoccurrenceofacroweddisaster.Basedonthisfinding,wecarriedoutanalysiswithfocusonthe“venuesuitability”whichshouldbedeterminedduringtheeventplanningstage.Theresultoftheanalysisshowedthatthevenuesuitabilityfromtheviewpointofcrowdsafetydependsonthenumberofvisitors,thetypeandcontentoftheeventaccordingtothevenuespaceutilizationplan,andthecrowdmovementintheaccess routes to and from the venue.

HYPOTHESISREGARDINGVENUESUITABILITYThesuitabilityofavenueforanygiveneventisthoughttodependonthenumberofvisitorstothevenue.Thevenuespaceutilizationplanwoulddeterminecrowdmovementwithinthevenue.Thecontentoftheeventmayaffectthecharacteristicsoftheeventandthetypes/agesofvisitors.Thetypeoftheeventmayaffectthecrowdmovementwithinthevenueandrequiredvenuespacecapacity.Thecontent of the event would determine the crowd movement to and from to the venue and within the venueduringtheevent.Unlessthesecrowdmovementsaresmooth,itishighlyprobablethattherewould be excessive accumulation of people.

CASESTUDY1Acrowddisastercausedbytheincorrectlypredictednumberofvisitorsandthecrowdmovementinanaccess route to the venue

1)Overviewoftheinterruptedhigh-densitycrowdflowThe“JapanCountdown2001”washeldatOkuraBeach,AkashiCity,fromtheeveningofDecember31,2000tothedawnofJanuary1,2001.Duringthedramaticlightilluminationshowandfireworkdisplay,aninterruptedflowofhigh-densitycrowdoccurrednearthesouth-sideendofthepedestrianoverpassconnectingtheeventvenuetothenearestrailwaystation.Theestimatedcrowddensitywas10to12peoplepersquaremeter,whichwasonthevergeofastampede.Urgentsecuritymeasuresweretakento

successfullyavoidastampede.Sevenmonthslater,however,asimilarinterruptedflowofhigh-densitycrowdoccurredonthesamepedestrianoverpass,causingafatalstampede,duringthe32ndAkashiCitizens’SummerFestivalheldatthesamevenue.

2)FactorsaffectingthevenuesuitabilityThenumberofvisitorspredictedbytheeventorganizerwas25,000,butactualnumberofvisitorswas55,000,2.2timesmorethantheprediction.Themethodforpredictionhadasignificantdefect;theorganizerdidnottakeintoaccountthenearbyresidentswhowouldcometothevenue.

Asforthecrowdcontrolplanintheaccessroutestothevenue,ifthenumberofvisitorswas25,000aspredictedbytheeventorganizer,itwouldbepossibletoleadthereturningcrowdthroughtheaccessroutesinoneandahalfhours.Ifthenumberofvisitorswas55,000,ontheotherhand,thereturningcrowdofapproximately44,000peoplewouldcomefloodingonthepedestrianoverpass,causinganinterruptedflowofhigh-densitycrowd.Topreventthis,itwouldbenecessarytosetacrowdtrafficcontrolline3minwidthand944minlength.Inaddition,thetimetakenfortheentirecrowdtopassthroughthepedestrianoverpasswasestimatedtobethreehoursintheory.Crowdcontrolundersuch conditions was unrealistic considering that the event was held at midnight on a new year’s day. Apparentlythevenuewasnotsuitableforsuchanevent.

CASESTUDY2Acrowddisastercausedbytheincorrectlypredictednumberofvisitors,inadequatevenueutilizationplan,andcrowdmovement

1)OverviewofthecrowddisasterAtthe2010LoveParadefestivalheldinDuisburg,Germany,acrowdofpeopletotheparadevenueranhead-onintotheon-comingcrowdreturningfromthevenue,causinganinterruptedflowofextrahighdensitycrowdandoccurrenceofacomplexcriticalcrowdwave.Twenty-onepeoplewerekilledandover500peoplewereinjuredduetocrowdpressureandfall.

2)FactorsaffectingthevenuesuitabilityThenumberofvisitorspredictedbytheeventorganizerwasonemillion.Thepoliceaskedtheorganizertosubmitthedocumentvalidatingtheprediction.Theactualnumberofvisitorswasestimatedtobe480,000basedonthepastrecords.Theeventisatypicalmixed-typeexcitingmusicfestivalattractingyoung people.

Thefestivalfeaturesstagemusicbyandfloatswithmusicmovinginfrontoftheaudience.Theeventvenuewassandwichedbetweenanexpresswayontheeastsideandarailwayonthewestside.Sincethenorth-sideaccessroutewaslimitedforusebytheinterestedpartyoftheevent,visitorscouldonlyusethesouth-sidemainaccessroadandsubaccessroad.Thevenuewasthereforeabottle-shaped,confinedarea.

CONCLUSIONPreventionofacrowddisasterinthevenueofalarge-scaleeventbeginswiththecheckupofvenuesuitabilityduringtheeventplanningstage.Itisessentialthattheeventorganizer,eventplanner,securitycompanyemployedbytheorganizer,police,fireagency,transportationcompanyandallotherconcernedpartiesshareinformationandmakediscussionsfromcomprehensivepointofview.

Poster exhibition

Unsorted list

Rainer Könnecke, IST GmbH, Frankfurt am Main GERMANY Effect of social groups on crowd dynamics - empirical findings and numerical simulations

Michael Seitz, Technische Universität München, München GERMANY Modelling pedestrian group behavior in a cellular automaton

Manuela Di Mauro, Earth Observatory of Singapore, Singapore SINGAPORE Pedestrian-vehicles interaction during evacuation: agent-based hybrid evacuation modelling of Southeast Asian cities

Marijn Swenne, Leiden University, Leiden NETHERLANDS Optimizing pedestrian environments with evolutionary strategies

Nese Cakici Alp, Gebze Yüksek Technology Institute, Gebze TURKEY Occupants emergency behaviour in turkey

Hubert Klüpfel, TraffGo HT GmbH, Flensburg GERMANY The Loveparade tragedy 2010 – causes and consequences

Amir S. Sahaleh, Swiss Federal Institute of Technology, Lausanne SWITZERLAND Microscopic calibration and validation of pedestrian models: Integrating discrete choice model into social force model Mohcine Chraibi, Forschungszentrum Jülich GmbH, Jülich GERMANY OpenPedSim: A framework for pedestrian flow analysis

Dirk Durst, Feuerwehr Kerpen, Kerpen GERMANY Large-scale multi-modal evacuation analysis with an application to Hamburg

Effect of social groups on crowd dynamics - empirical findings and numerical simulations

Rainer Könnecke, IST GmbH, Frankfurt am Main GERMANYDirk Oberhagemann, VFDB e.V., Lippetal GERMANYVolker Schneider, IST GmbH, Frankfurt am Main GERMANY

Majorlargescaleeventslikepublicviewing,openairconcertsorlargefestivalsmakehighdemandsonallresponsibleparties,especiallywithrespecttohumansafety.Fromdisastersthatoccurredinthepastitcanbeconcludedthatcrowdincidentswithseriousconsequencestolifeandhealthofpeoplecanhaveavarietyofcauses.Soitisnotpossibletofocusonemergencysituationsalone,butitisrequiredtoanalysealargenumberofscenariosintheforefrontofsuchlargescaleevents.Usually,suchscenarioswillcomprisehighdensitycrowdmovement.Thelackofinformationonlimitingfactorslikenumberofvisitors,availablespaceandbehaviouralaspectsmakesitdifficulttodefineandassessscenariosforreliablesafetyconcepts.HenceajointresearchprojecttitledEVA(RisikoGroßveranstaltungen–Planung,Bewertung,EVAkuierungundRettungskonzepte)hasbeenestablishedinordertodefineandanalyzerelevantdesignparameters.TheEVAprojectstartedinMarch2009andwillbefinishbyFebruary2012.ThispaperwillfocusonempiricaldatacollectionandanalysisanditsuseforthecalibrationandvalidationofthemicroscopicevacuationandpedestriandynamicsmodelASERI.ThefacetofthetotalEVAprojectpresentedherewascoveredbytwooftheprojectmembers,theGermanFireProtectionAssociation(vfdb)andISTGmbH. Alargeamountofdata-mainlyintheformofvideorecordings-wascollectedbyDirkOberhagemannandhisvfdbteam.FurtherempiricalinputwasprovidedbytheDisasterResearchAgency-InstituteforSociologyoftheChristianAlbrechtsUniversityinKielandtheInstituteofFireandEmergencyTechnologies(IFRDortmund).Whilethedataanalysisfromthevfdbgroupwasessentiallymanually,theFraunhofer-InstituteforChemicalTechnologyICTcontributedcomputerizeddigitalization,filteringandvisualizationtechniques.

Internationaldesignparametersforassemblyareasarestronglyvarying,includingvaluesaslowas0.7persons/m2(Italy)upto4persons/m2(Switzerland).InGermanyanumberof2persons/m2isapplied.Insomecountries,thesedesignparametersareformallyrestrictedtobuildings.ThevfdbteamhasmeasuredstaticanddynamicdensitiesofpedestriansatvariouseventsinGermanyandhasfurthermoreanalysedavailablearchivematerial,e.g.frompublicviewingareasduringtheFIFAWorldTeamCup2006.Itwasthuspossibletocollectalargeamountofdataonpeakandsustaineddensitiesespeciallyindynamicsituations.Thisdatawasfurtheranalysedwithrespecttotheformationandmovementofsocialgroups.Usually,themajorityofvisitorsoflargescaleeventsaremembersofsocialgroupsofsizevaryingbetween2and6persons.Byidentifyingthemovementofsuchgroupsfromthevideosequencestypicalmovementandbehaviouralpatternscouldbederiveddependingongroupsizeandsocialfactorslikegenderorparent-childrelations.Thisbehaviourwasclassifiedaccordingtoaparameterdescribingthetendencytoconservethecompactnessofthegroupduringmovement,withsmallergroupsbeingmoreoffensivethanmediumorlargesizegroupsandfamiliesbeinglessoffensive

thangroupsoffriendsorcasualacquaintances.Itwasfurthermorepossibletoestablishfundamentaldiagramsofwalkingspeedandflowversusdensityforvariousgroupsizes.Theeffectofbarriersongroupmovementwasalsoinvestigated,howeverwithoutrevealingdistinctfeatures.Itwasthusconcludedthattheimpactofbarriersismorecloselyrelatedtospecificenvironmentalfactors,whilegroup movement in general is strongly determined by social relations.

Microscopic modelling is a very fundamental and thus powerful method to investigate crowd movementincomplexsurroundings.Thisapproachisabletocoverconsistentlylowandhighpedestriandensityregimesandcanbeappliedtoverylargebuildings,facilitiesandopenareas.WithintheEVAresearchprojectthemicroscopicmodelASERIdevelopedbyISTGmbHwasextendedtoapplicationstypicalforlargescaleeventsinemergencyandnon-emergencysituationsbymodifyingandextendingthebasicmovementandbehaviouralsub-modelsandperformingtherespectivecalibrationandvalidationwork.Thisincludescrowdmovementatveryhighoccupantdensitiesandlargetotalpopulationnumbers,counter-flow,laneformationandthecollectivemovementofsocialgroups.

ItisnowpossibletospecifythenumberofgroupsofagivensizeaspartofascenarioforanASERIsimulation-eitherforanemergencysituationlikeevacuationcausedbyimmediatedangerornon-emergencysituationslikecrowdmovementathighpopulationdensity.Basedonthespecifiednumberandsizeofgroupsappropriateinitialdistributionsofgroupsaregeneratedanddistributedwithintherespectivestartingareas.Thegroupvelocityisgovernedbytheminimumunrestrictedwalkingspeedof the individual group members. In order to model the dynamics of group coherence in a crowded environment,socialaffiliationisanimportantfactor.ASERIincludesfourtypesofinterpersonalrelations:mutualpositive,mutualnegative,asymmetric(e.g.parent-childcombinations)andneutral(nospecificsocialrelation).Theserelationsdefinetherangeofmutualinterpersonaldistancesthatare tolerated without violating the coherence of the group. If due to the movement of other people inacrowdedsituationtheinterpersonaldistanceexceedstherespectivetolerablelimit,thegroupmembersfurtheraheadwillslowdownorevenwaituntilgroupcoherenceisagainestablished.Usingtherelativedistributionofgroupsizeandsolitarypedestriansfromtheobservationofthevfdbteamthefundamentaldiagramiscalculatedfordensitiesbetween0.4persons/m2and2.4persons/m2.Thefollowinginterpersonalrelationswereassumedfortherespectivegroups:50%oftherelationstoothergroupmembersareofthetype„mutualpositive“and50%ofthe„neutral“type.Theminimuminterpersonaldistancetoagroupmemberis1mfortype„mutualpositive“and5mfor„neutral“.Withthis set of parameters the empirical fundamental diagrams could be reproduced with good agreement.

Futureworkwillincludeaspectsoflaneformationverycrowdedsituations.Recentobservationshavedocumentedoscillatingeffectsinlaneformationandevenmutualbackwardflowatveryhighoccupantdensities.

Modelling pedestrian group behavior in a cellular automaton

Michael Seitz, Technische Universität München, München GERMANYGerta Köster, Technische Universität München, München GERMANYAlexander Pfaffinger, Siemens AG, München GERMANY

Thestudyofcrowdmovementisveryimportantforplanningmasseventsandevacuations.Althoughmanypotentiallycriticalincidentsarehardlypredictable,otherslikeclogging,canbeanticipated.Studieshaveshownthatpedestriansinsocialgroupsfrequentlycontributethebiggestparttocrowds(Aveni,1977;Singhetal.,2009),andthatthegroupshaveasignificantimpactonthecrowdmovement(Moussaidetal.,2010,Kösteretal.,2012).Furthermoresocialcooperativebehaviourdoesnotstopinemergencysituations,butcontinuesorevengetsstronger(Druryetal.,2009).However,mostofthe currently successful simulation models that are able to capture important characteristics of crowd movement,likedensity-flowrelationsortheformationoflanes,relyonmodelsfromphysics,suchasNewtonianmechanics,electrodynamicsorfluiddynamics.Wearguethatthemodellingofindividualpedestriansintheformofe.g.particlesisnotsufficienttocapturecrowdmovement.Inthecourseofourmodeldevelopmentwefoundthatthereisaneedforagent-basedmodellingor,atleastasinourcase,agent-typemechanismstosupplementthephysicallyinspiredmodel.

Themodelofindividualbehaviourweuseforoursimulationisacalibratedcellularautomatonthatisabletosimulatescenarioswithahighnumberofpedestriansatthesametime(Kösteretal.,2010).Wepresentabasicmodelforthemovementofsocialgroupsincrowds.Thefirstgoalistomaintaingroupcoherenceforgroupsofsizetwotofourinarealisticwaywhilethecrowdmovestoagiventarget(Kösteretal.,2012).Othermodelshavebeenabletoachievethis(Singhetal.,2009;Moussaidetal.,2010)andwediscusssimilaritiesandwhytheydonotfitourneeds:Inourmodel,wefirstintroduce an attractive potential around the member of the group who is most advanced towards the target.Secondlyweletmembersthatarefallingbehindspeedupandmembersthataregettingaheadslow down. Furthermore we intoduce the concept of a group leader as a point of orientation for the othergroupmembers.Singhetal.(2009)alsousealeader,butintheirmodeltheleaderisafixedperson,wherasinourmodeltheleaderrolecanbepassedon.Moussaidetal.(2010)usethecenterofmassasattractionpointforallmembers.Byanalyzingvideofootagetheycometotheconlusionthat microscopic behaviour can be explained by the urge to communicate. Mousaid’s mathematical formulationistargetedforimplementationinasocialforcemodel.Itusesaccerlationanddeceleration,whereasthecellularautomatondirectlycomputesthedirectionandspeedatagivenmoment.Hencewe model the communication phenomenon in a way suitable for the cellular automaton approach.

Wecalibratethemodelaccordingtoasmallevacuationexperimentweconducted(Kösteretal.2012).Duetothedirectattractiontotheleader,pedestrianscangetstuckwhentheyareseparatedbyanobstacle.Thustheseparationofgroupsbyanobstaclehastobedetectedandresolved.Onewaytodothis,istoletthelostmemberstrackbackinordertofindthegroupagain(Seitzetal.,2011).Anothersolutionistousegraph-basedrouting.

Atmasseventsgroupsofbiggersizearetobeexpected.Acommonassumption,yetwithoutempiricalevidence,isthatlargegroupsseperateintosmallersubgroupsofsizetwotofour(Moussaidetal.,2010).Weadditionallyassumethatlargegroupsalsostaytogetherasawholeinalessstrictformation.Nocommunicationbetweenmembersoftwodifferentsubgroupsisassumed.Otherwiseweusethesamemechanismweusedforsmallgroupsbutwithweakerinfluenceonthemovement.Themodelforthesubgroupsstaysidenticaltotheoneforsmallgroups.Simulationrunsimplyastrongimpactoncrowdmovement in certain scenarios.

Thediscussedmechanismscanbedividedintwocategoriesofdecisionmaking.Thefirstapproachcombinesdifferentinfluences,e.g.targetattraction,repulsionfromotherpedestrians,attractiontotheleaderandchoosesthebestcompromise,mostlikelybyaddingupthepotentials.Thisworksaslongasthesituationissomewhatsmooth.Buttherearesituationswherethereisnoreasonablecompromiseandoneormoreinfluenceshavetobeneglectedcompletely,asitisthecasewhenagroupmemberhasbeenlost.Thenthepedestrianhastodecidewhethertocontinuewalkingtowardsthetargetortrytofindthegroup.Werealizethisbinarydecisionbyintroducingagent-typemechanismsintothemodel.Weconcludethattheabilitytotakedecisionsthatcannotbeexpressedthroughaphysicallyinspiredsuperposition of forces is necessary when it comes to more complex social pedestrian behaviour.

REFERENCES

Aveni, A.F. (1977) The Not-So-Lonely Crowd: Friendship Groups in Collective Behavior, Sociometry, 40(1), 96-99.

Drury, J., Cocking, C., & Reicher, S. (2009) Everyone for themselves? A comparative study of crowd solidarity among emergency survivors, British Journal of Social Psychology, 48, 487-506.

Köster, G., Hartmann, D., & Klein, W. (2010) Microscopic pedestrian simulations: From passenger exchange times to regional evacuation, in Proceedings of the International conference on operations research: mastering complexity, Munich, Germany, 1-3 September 2010.

Köster, G., Seitz, M., Treml, F., Hartmann, D., & Klein, W. (2012) On modeling the influence of group formations in a crowd, Journal of Contemporary Social Science, accepted.

Moussaïd, M., Perozo, N., Garnier, S., Helbing, D., & Theraulaz, G. (2010) The Walking Behaviour of Pedestrian Social Groups and Its Impact on Crowd Dynamics, PloS ONE, 5(4).

Singh, H., Arter, R., Dodd, L., Langston, P., Lester, E., & Drury, J. (2009) Modelling subgroup behaviour in crowd dynamics DEM simulation, Applied Mathematical Modelling, 33, pp.4408-4423.

Seitz, M., Köster, G., & Hartmann, D. (2011) On Modeling the Separation and Reunion of Social Groups, in Proceedings of the International Conference on Emergency Evacuation of People from Buildings, Warsaw, Poland, 31. March - 1. April 2011.

Pedestrian-vehicles interaction during evacuation: agent-based hybrid evacuation modelling of Southeast Asian cities

Manuela Di Mauro, Earth Observatory of Singapore, Singapore SINGAPOREMichael Harold, Nanyang Technological University, Singapore SINGAPOREKusnowidja Megawati, Nanyang Technological University, Singapore SINGAPOREZhenhua Huang, Nanyang Technological University, Singapore SINGAPORE

Inmodellingmassevacuation,vehiclesandpedestrianareusuallymodelledseparately:allowingforsometransversalinteractions,tosimulatepedestriancrossingtheroads;and/orlongitudinalinteractions,byintroducingsomedegreeofreductionintheroadcapacitywhenpedestriantemporarilyhopoffandonthefootpath(e.g.MeschiniandGentile,2011;WuandWu,2010).Suchmodelsassumethatpedestrianswillmainlyremainwithinthedelimitedfootpathduringanevacuation.However,inmanycities,particularlyinSoutheastAsia,pedestrianandvehiculartrafficflowsarenotrigidlyseparated,eveninnormal(non-emergency)situations,seeforexampleChandraandKumar(2003).Inmanycases,urbanroadsdonothavefootpaths,andbothvehiclesandpedestrianstendtooccupyanypossiblefreespaceintheroads.Thiscouplingofpedestrianandtrafficdynamicswillgreatlyimpactanylargescaleevacuationinsuchcities.Tomodelthisformofevacuationitisthereforenecessarytoaccuratelycapturetheimpactwhichtraffichasonpedestriansand vice versa.

Thispaperpresentstheresultsofastudyaimingtodevelopahybridevacuationmodelabletotargettheseissues,inordertobettersimulatehybridpedestrianandvehicularevacuationinSoutheastAsiacities.Thepaperpresentsanovelmodelthatisabletosimulatetheinteractionbetweenvehicleandpedestrians,alsocapturingthepeculiarconditionsofthetrafficinmanySoutheastAsiancities.Infact,mostofthecommonlyusedmodelsfortraffic(bothmacroandmicroscale)aredesignedforspecificorderlytraffic,assumingthatthevehiclewouldfollowqueuingpatternsorlanedivisions,featuresnotapplicabletomanycitiesinSoutheastAsia,whichalsopresentaveryspecifictrafficcomposition,includingahighrateofmotorbikesandsmallervehicles.Themodelisdesignedtobescalableandisable to handle a large number of evacuees. Theresultingmodelisanagent-basedmodelthatsimulatespedestrianandvehicles,usingthejava-basedmodellingplatformMASON.Thevehicle’smovementismodelledusinglocallysourcedempiricalfundamentaldiagramstorepresenttheflowsattheroadlinksandintersections.Ateachtimestep,thevehiclesmovealongtheroadlinkswithaspeedthatisfunctionoftheflow(assumedconstantduringthetimestep)andthecapacityoftheroad.Similarly,empiricalfundamentaldiagramsfortheintersectionsareusedtomodeltherelationshipsbetweenthecapacityandactualinflow,whichdeterminetheeffectiveinflowandwaitingtimeattheintersections.ThepedestrianmovementissimulatedusingWeildmann’sdiagramforpedestrianalongthelinksandtestingdifferentalgorithmsformodellingthedelayattheintersections,includingempiricaldiagramsforpedestriannetworks(e.g.Hoogendoornetal,2011).Thecapacitiesofroadsandintersections,forbothvehiclesandpedestrian,vary at each time step as functions of the total number of agents (pedestrians and vehicles) and the ratio pedestrian\vehicles.

ThefirstpilotsiteforthismodelisthecityofPadang,Indonesia,whichwaschosenforseveralreasons.First,thecityisexpectedtobeaffectedbyanearthquakeandsubsequenttsunamiinthenextdecades.Althoughevacuationstudieshavebeencarriedoninthepast,ourconsultationswithlocal emergency planners showed the need for more detailed analysis and assistance in translating themintoactualevacuationplans.Secondly,thecitywasrecentlysubjecttoamajorearthquakethattriggeredamassevacuation,whichshowedthetypicaltrafficcharacteristicsobservedforSoutheastAsiancitiesdescribedabove.AnotherreasonforchoosingPadangaspilotsiteistheexistenceoftwomodelslookingatthecity’sevacuationthatcanbeusedforcomparison:aqueuingmodeldevelopedbyLammeletal.(2010)andaGIS-basedmodeldevelopedbyPostetal.(2010).ThelatterisaGIS-based tool that statically calculates possible evacuation times based on the physical distances to safe places,landusesandpopulationcharacteristics.ThefirstisapedestrianmodeldevelopedbyadaptingaMatSIMqueuingvehicularmodel,usingWeidmann’sfundamentaldiagramandassigninga‘storagecapacity’forsimulatingthedelayatthenodes.Thismodelisabletosimulatealargenumberofpedestrians,andembedsalgorithmsthatcalculateoptimalorpseudo-optimalsolutions.Thismodel,however,includeonlypedestrians’movement.Also,lookingatthe‘optimal’solutionsdoesnotaccountformorecomplexscenarios.Inthissense,themodelproposedinthispaperisdesignedtoembeddifferentroutechoices,whichwillbeeventuallytreatedasprobabilisticensembles.Thispaperpresentsthepreliminaryresultsoftheapplicationofthismodel,bytestingthesensitivityoftheresultstodifferentfundamentaldiagramsandroutechoices,andcomparingtheresultswiththeabovemodels,andwithotheravailablenon-hybridpedestrianandvehicularmodels.ThisapplicationwascarriedonthankstothecollaborationwiththelocalemergencyplannersandNGOs.Oncerefined,themodelwillalsobeappliedandfurthervalidatedtootherSoutheastAsiancities.

REFERENCES

Chandra, S. and Kumar, U., 2003. Effect of Lane Width on Capacity under Mixed Traffic Conditions in India. Journal of Transportation Engineering, 129, 2, 155-160

Hoogendoorn, S.P., Campanella, M.C. and Daamen, W., 2011. Fundamental Diagrams for Pedestrian Networks. In Peacock, R.D. et al. (eds.), Pedestrian and Evacuation Dynamics 2010, Springer, 255-264

Lammel, G. et al. 2010. Emergency Preparedness in the Case of a Tsunami— Evacuation Analysis and Traffic Optimization for the Indonesian City of Padang. In Wolfram, W.F. et al. (eds), Pedestrian and Evacuation Dynamics 2008, Springer, 171-182

Meschini, L., Gentile, G., 2011. Simulating car-pedestrian interactions during mass events with DTA models: the case of Vancouver Winter Olympic Games. Proc. of 90th annual meeting of the Transportation Research Board, Washington D.C.

Post, J. et al. 2009. Assessment of human immediate response capability related to tsunami threats in Indonesia at a sub-national scale. Natural Hazards Earth System Science, 9, 1075–1086

Wu, H. and Wu, W., 2010. Microscopic Dynamic Simulation Model for Pedestrian-Vehicle Mixed Traffic. Proc. of IEEE International Conference on E-Health Networking, Digital Ecosystems and Technologies

Optimizing pedestrian environments with evolutionary strategies

Marijn Swenne, Leiden University, Leiden NETHERLANDSThomas Bäck, Leiden University, Leiden NETHERLANDS

INTRODUCTIONMany pedestrian crowd simulators have attempted to emulate or study their behaviour for the purposeofpredictinghowpedestrianswouldbehaveinvariousenvironments.Amongstthemostpopular of these settings are the evacuation scenarios where all pedestrians have to exit a building assoonaspossible.Optimizingthisenvironmenthastraditionallybeentheworkofarchitectsorotherprofessionals.Weproposeamethodforoptimisingthisgeometricenvironmentbyuseofanevolutionarystrategy.Simulatorpredictionswilldeterminethedesirabilityorfitnessoftheseenvironments.ThesimulatorwebuildforcalculatingthefitnessisbasedonHelbing‘ssocialforcesmodelandpointofvisibilitypathfinding.Wewillshowthatthisapproachgeneratessurprisingresultsand can generate geometries which help decreasing evacuation time.

PEDESTRIANPATHPLANINGInorderfortheagenttoreachhisgoal,hewillavoidallobstaclesinamosthumanlikefashion(asapposedtothemostefficientway).Theseobstaclescanbeclassifiedasstationaryorstaticobstaclesanddynamicobstacles.Thegeometricenvironmentisconsideredtobeasetofstaticobstacleswhichwillbe circumnavigated mainly with use of a visibility graph. Fellow agents will be considered as dynamic obstaclescircumnavigatedbyusingthesocialforcesmodel[1].

Betweenstaticobstacles,whicharerepresentedaspolygons,avisibilitygraphisconstructedrepresentingpossibleglobalpathsthatagentscanfollow.Agentschoosethepathwiththeshortestestimatedtraveltime,whichinturnisbasedonlengthandcongestionrates.Fromthispathanattractorpointneedstobeselectedforeachagenttocalculatethedirectionofitsnextstep.Commontechniques[2,3]don‘tworkwellformulti-agentscenariosinnarrowenvironments.Wedesignedamethod that creates a polygon around an agent with the attractor point being the intersection of the polygonedgeswiththepathorthelastpointonthepathifnointersectionispresent.Thepolygonisconstructedbycreatingperpendicularlinesforeachobstacleonit‘sclosestpointinrelationtotheagent.Thencreatinglinesegmentsbycalculatingtheintersectionsoftheselinesandconcatenatingtheinner line segments to construct a convex polygon.

Toprocessdynamicobstacles,thesocialforcesmodelwouldnormallytakeallfellowagentsintoaccountandhavethemprojectvectorsontheagentrepresentingasocialforces,pushingtheagentawayfromthedynamicobstacle.Themagnitudeisrelatedtothedistancebetweentheagentandobstacle.Thiscreatessomebutnotalloftheemergentbehaviourseeninpedestrians[4].WeaddedasecondsetofcriteriabasedonGoffman‘sresearch[5]onhowpedestrianstakenoticeofotherpedestrians.Bothadditionsgiveusamodelthatismoretruetonatureandshowsbetterpathfindingandbetterlane

formation in dense streams.

GEOMETRICENVIRONMENTOPTIMIZATIONWithourgeometricenvironmentbeingacollectionofpolygons,wewanttoaddorrearrangetheseinordertomaximisethecapacityorthroughput.Toautomatethisoptimizationprocess,wewillmakeuseofaheuristicoptimizationtechniqueknownasanEvolutionaryStrategyorES[6].

Aheuristicapproachwaschosenbecauseitisabletoproduceagoodsolutioninapracticalamountoftime.TheESinparticularwaschosenbecauseitoperatesinreal-valuedsearchspaceandsuitsthegeometricenvironmentwhichisinacontinuessearchspace.Givenafitnessfunctionandasetofreal-values,anESwilltrytofindasetofvaluesthatproducesthebestfitnesswithtimeincreasingtheoddsoffindingabettersolution.Inthisparticularcase,therealvaluednumbersrepresentstaticobstacles.Weuseatranslationwheresomenumbersareusedtorepresenttheshapeandsometorepresentthelocation of the obstacle.

EXPERIMENTALSETUPWeconstructedascenarioconsistingoftwoparallelwalls,makingacorridoranddictatingthegeneralflowoftheagents.PillarsareaddedbetweenthesewallsofwhichthesizeandlocationarerepresentedasrealvaluednumberstobeoptimisedbytheES.Agentsenterthescenarioonoppositesidesofthecorridorandhavetheirgoalsettotheoppositeendsoftheirstartinglocation.Throughputismeasuredbymakingasummationofallstepsneededforeachagenttoreachitsgoal,wherefewerstepsequalsabetterfitness.Asacontrolexperimentwecreatedthesamecorridorwithoutpillarsandusedthesimulatortocalculateit‘sthroughput.

RESULTSResultsshowthatwhenplacingpillarsinacorridor,certainarrangementsexistwhosethroughputdoesnotvarysignificantlyfromhavingnopillars.Someoftheselocaloptimahavepillarsplacedinthemiddleofthecorridor.Thesepillarsseemtosplitagentsintostreamsandmakingthemavoidfuturecollisions.Currentlywearetestingadditionalscenarioswiththegoalofincreasingthroughputbyeitherinsertingobstacles(fences/pillars)orbyrearrangingexistingobstacles(seating/smallshops).Preliminaryresultslookpromisingandwillbeavailableforthefullpaper.

OUTLOOKFutureresearchwillconsistofextendingtheagentswithapsychologicalmodelbasedontheOCEANmodelandintroducingbodyfactorssuchasweight,compressibilityandinjury.

REFERENCES

[1] Proulx, G. (1995). Evacuation Time and Movement in Apartment Buildings. Fire Safety Journal, 24, 229-246.

[2] Purser, D. (2001). Quantification of behavior for engineering design standards and escape time calculations. Safety Science, 38(2), 157-182.

[3] Santos, G.; Aguirre, B. E. (2004). A Critical Review of Emergency Evacuation Simulation Models.

[4] Kuligowski, Erica D. (1999). The Process of Human Behavior in Fires. NIST Technical Note 1632.

Occupants emergency behaviour in turkey

Nese Cakici Alp, Gebze Yüksek Technology Institute, Gebze TURKEYGülen Çagdas, Istanbul Technical University, Istanbul TURKEY

Duringemergencyinhighpopulatedbuildings,likehigh-rise,stadium,theaterandhospitalshumanlifesafetyisveryimportant.Thispopulatedbuildingsoccupantsmustbeprotectedbybuildingsdesigntactics.Foramoreeffectivecirculationsystemdesigninthebuilding,environmentalfactorsandoccupantcharacteristicsshouldbeexaminedthoroughly.Forthatreasonsince1970’stheaimofthescientificfireengineeringandmostofotherdisciplineslikearchitectureresearchesfortounderstandhumanevacuationbehaviourunderemergencysituationinbuildings.Alsoeverysocietyhasitsowncharacteristics.Humanpopulationshavesomanydifferencesinrespectofphysiologic,anatomic,behaviouralandanthropometricviews.Thosedifferencesoccurforthereasonoftheinteractionofgeneticnature,environmentalfactorsandculturallivingstandardsofthesociety.Duetothispoint,usingTurkishhumananthropometric,behavioraldataisefficienttorealisticoutputofevacuationanalysisinTurkey.ForthispurposeasurveyaskedtooccupantsoftwodifferentbuildingsSurveyanswersanalyzedwithSPSSstatisticalsoftware.DispersionsgivenbysoftwarewhicharedeterminingtheemergencybehaviourofoccupantscomparedthedispersionsofU.S.A.andEnglandoccupants’characteristic’stounderstandculturalbehaviouraldifferencesofsocieties.Howevermorebuildingsandtheiroccupantsneededtoanalyzed.

Surveyswereconductedintwodifferentbuildingforunderstandingoccupant’semergencybehaviourandpsychology.Surveysofthedifferentstudiesonthebehaviorofanemergencysituation(Wood,1972,FahyandProulx1995;Sekizawaetal.,1999,ProulxandReid,2006)examinedandevaluatedpriortostudysurveywasdevelopment.Themainpurposeofanalyzingthesesurveysistoinvestigateoccupant’s emergency behaviour and to understand the dispertion of emergency behaviour changes in differentoccupantsgroup.

Thefirstofthesesurveyswasadministeredon01.04.2010followingthefirethatbrokeoutonthegroundflooroftheFacultyofArchitectureoftheGIT(GebzeInstituteofTechnology),thearchitecturalplanofwhichisshowninFigure1.a,onoccupantswhowereinthebuildingatthetimeofthefire.Atotalof27peopleworkatthefacultyofarchitecture,whichconsistsofagroundfloorandafirstfloor;inaddition,graduatecoursesaregiveninthebuildingatsomehoursoftheweek.Atthetimeofthefire,therewereatotalof17peopleinthebuilding.Thesurveyquestionswereposedtothese17peoplewhowereinthebuildingduringthefire.ThesecondsurveywasadministeredontheemployeesoftheKocaeliMetropolitanMunicipality(KMM)Building,thegroundfloorplanofwhichisshowninFigure1.b.Thebuildingwhere600areemployedconsistsof2towers,onewith5andtheotherwith6floors,mergedwitha2-storeymainbuilding;thequestionnairewasadministeredon123peoplefromamongtheoccupantsofthegroundfloor,the1stfloorandthe2ndfloor.

Thepurposeofthesurveywastostatisticallyidentifytheemergencybehaviourandpsychologyoftheoccupantsandthestatisticalbreakdownofthe‘duringfire’behaviouraspertheoccupants’genders,theirknowledgeaboutthebuilding,andtheirpositioninthebuilding.ThevariablesusedintheanalysisofthesurveyaregiveninTable1.Asindependentvariablesarenotmetric,theKruskal-Wallistestwasemployed,whichconductsaone-wayvarianceanalysisbetweengroups.

Tableofdependentandindependentvariables:

Dependentvariables Independentvariables• Occupant’sage• Occupant’ssex• Occupant’sworkingfloor• Occupant’sdurationtime • Responsetofirealarm• Cueforfireperception• Firstreactionafterfireperception.• Evacuationspeed• Reactionafterfacingtosmoke• Firstexitroutechoiceinfire

Fromthesurvey,itisunderstoodthatoccupantsdemonstrateatendencytousetheemergencyexitdoors,yetfailtoadequatelypicturethebuildingintheirminds.Amonggroundflooroccupants,theratioofthoseusingtheemergencyexitdoorsandthoseusingthewindowsareequal.Unliketheoccupantsofupperfloors,itisstatisticallyproventhatgroundflooroccupantsaremorelikelytousethewindowsforexitpurposes.ForTurkishpeople,itisseenthattheemergencybehaviourisnotrelatedtooccupantage,butrelatedtogenderandthefloorwheretheoccupantislocated.Thefloorwheretheoccupantislocatedaffectstheoccupant’sescapeperformance.Itisunderstoodthattheescapetendencyofupperflooroccupantsislowercomparedtolowerflooroccupants.Whencomparedtoothercountries,theemergencybehaviourofTurkishpeopleappearstodifferintermsofprioritiesandbreakdowns.Inthesurveys,theemergencyprofilesoftheoccupantsweredrawnthroughanalysisofalimitedsample.Expandingtheseemergencyprofilescountrywidewouldincreasetheconsistencyoftheprofiles.Occupantbehaviourtrendsanddistributionsalsovaryconstantlyandinthecourseoftimewiththedevelopmentsandchangesinbuildingstructuresandinlifestyles.Hence,itwouldbeappropriate to conduct similar surveys across the country and on a permanent basis.

REFERENCES

Bryan J., 1995: Behavioural Response to Fire and Smoke, the SFPE, Handbook of Fire Protection Engineering2nd Edition, National Fire Protection Association, Quincy, MA, 3, 241-262.

Çakıcı N., 2011: Simulation And Representation Of Human Emergency Behaviour And Movement In Buildings Using An Agent Based Model, PhD Thesis, Istanbul Technical University, Istanbul, Turkey.

Fahy, R.F., Proulx G., 1995: A Study of Human Behaviour during the World Trade Center Evacuation, NFA Journal, 59-67.

Proulx G., Reid I. M. A: , 2006: Occupant Behaviour and Evacuation during the Chicago Cook County Administration Building Fire, Journal of Fire Protection Engineering, 16, 283-309.

Sekizawa A., Ebihara M., Nakote H., Kubota K., Nakano M., Ohmiya Y. and Kaneko H., 1999: Occupant’s Behaviour in Response to the High-rise Apartments Fire in Hiroshima City, Fire and Materials, 23, 297-303.

Wood G. P., 1972: The Behavior of People in Fires, Fire research note 953, Building Research Establishment, United Kingdom.

The Loveparade tragedy 2010 – causes and consequences

Hubert Klüpfel, TraffGo HT GmbH, Flensburg GERMANY

INTRODUTIONTheLoveparadewasanannualfestivalstartedinBerlininthe1990s.ItmovedtotheRuhrAreain2008Dortmund,2009Essen,2009Bochum–cancelled,2010Duisburg.

ThecitiesofDortmundandEssenhavearound600,000inhabitants,Duisburg500,000,Bochumaround350,000.In2009theeventwascancelledinBochum,mainlyduetosafetyconsiderations.Theofficialstatementwasthattherewasnotenoughroominthecitytohostsuchalargenumberofvisitors.

SOCIALANDPOLITICALCONTEXTTheLoveparadechangedfromBerlintotheRuhrAreain2008.Intheyearbefore,noLoveparadewasheldinBerlin.Thereasonwasthatthestatusof„politicaldemonstration“wasnolongergrantedandtherefore,thefestivalmanagementhadtopayforwasteremoval,trafficmanagement,etc.

LopaventorganizedkindofacompetitionbetweendifferentcitiesintheRuhrareaforbeinghostoftheevent.Itsabitcomparabletowhathappensforlargesportseventslikeolympicgamesorfootballworldchampionships(onasmallerscale,ofcourse).Atthesametime,theRuhrArea,whichwas–andstillis–theheartoftheGermansteelandcoalindustries,severelysuffersfromthestructuralchangesconnected to the decline of those industries.

THEDISCUSSIONINDUISBURGBEFORETHELOVEPARADEThediscussioninDuisburgmainlyfocussedonfinancialaspects.Sincethereisnoapprovedbudgetforthecity,expensescanonlybemadeiftheregionalgovernment(Bezirksregierung)approves.Sincethefestivalwascancelledintheyearbefore,therewassomepressuretofindasolutionforthefinancial„problem“,i.e.~findingsponsorstocoverthecostsforthecity.

THEVENUEThevenuewasaformerfreightrailstation.ItisveryclosetoDuisburgcentralstation.Theoverallsizeoftheareaisabout500,000sqm.OnlypartofitwasusedfortheLoveparade,though.SinceithasInthenorthofthevenueisthecentralstation,eastistherailline(10tracks)inthesouththereisanon-/off-ramptothehighwayA59whichisthewestborderofthevenue.Thereisanoldbuildinginthecenter of the area which was partially used during the day for screenings but not accessible to visitors. ThehighwayA59wasclosedduringthedayandmarkedasevacuationrouteandaccesspathforfirstresponders.

THEDAYOFTHELOVEPARADEThedaytheloveparadetookplacewasSaturday,July24th.Itwaswarmandsunnysummerday.Therewasadelaywhenopeningthevenueforthepublic.Theopeningwasscheduledfor10hrs,thedelaywasabout2hrs.Thereasonwasthattherewasstillworktobedonetopreparethevenue.

MostofthevisitorscameviaDuisburgcentralstation,whichisahubforICE,IC,andregionaltrainsaswellasmetroandbusses.Thewereguidedontodifferentpathstoatunnelandfurthertoanon-ramptothevenuelocatedinthesouthofthearea.Thefirstmajoreventwasscheduledat17hrs.At16hrs,severalproblemsconcerningtheinflowtothefestivalareaculminatedinastand-stillattheuperendoftheon-ramp.

Asecondramptothelocationaround40mwestofthemajorrampwasnotusedonthedayuntiltheearlyafternoonwhenthestand-stilloccured.

Thedensityontherampincreasedfurther,whenthestand-stillontheupperendoftherampcouldnotberesolvedandtherewasapolicelineontheramptryingtomitigatethesituation.Twomorepolicelinesattheentrancestothetunnelbrokedownwhichcausedfurtherpressure.Thepersonsontherampsawthreewaystoescapethesituation:Thecontainerofthecrowdmanageratthesouthernendoftheramp,alightpostontheeasternside-walland–mostofall–asmallstairatthewesternwalloftheon-ramp.Itwasthisstairwherepeoplesuffocatedanddied.

Thedetailsofthecrowdcrushhavebeencoveredextensively.Atthispoint,ahinttosomesourcesshouldsuffice(willbeputintothefullpaper).

THEMEDIACOVERAGEThemediacoveragewasextensive.Thedisasterswasmostlycalled„masspanic“.Forclarification,werefertothecontribution„LargeScaleEventsCrowdSafety“(Klüpfel,intheseproceedings).Itisimportanttostressthefollowingpoint,though:The„panic“isnotthereasonfortheinjuriesanddeaths.Peopledonotdie,becausethepanic.Theypanic,becausethedie.Thereasonforthepanicisthe extremely high density and forces and the imminent danger.

THELEGALCONSEQUENCESThereiscurrently(asofNovember2011)alegalinvestigationgoingon.Ithasnotyetbeenfinalized.ApreliminaryreporthasbeensubmittedtothestateparliamentofNorth-RhineWestphalia.

LESSONSTOBELEARNEDDifferentlevelsofanalysisanddetailarenecessarywhentryingtointerstandhowandwhythedisasterattheLoveparade2010happened.Therehavebeenmistakesintheplanning,thedesign,andduringtheday.Theaimofthispaperwasnottoidentifywhodidwhatwrongbuttosummarizethepubliclyavailableinformationandprovidereferencesforthereader.Shecanthendrawherownconclusions.Onelessontobelearnedmightbetohaveafocuson„softfactors“likepoliticalpressures,groupthink,selectiveperceptionandreasoning.Anotheronemightbethefocusonthesystemsapproachandtheidenficationofpositivefeedbackloops(likedensity→stress→noise→forces→density)whichmightbegeneratedduringaneventliketheloveparade.

Microscopic calibration and validation of pedestrian models: Integrating discrete choice model into social force model

Amir S. Sahaleh, Swiss Federal Institute of Technology, Lausanne SWITZERLANDMichel Bierlaire, Swiss Federal Institute of Technology, Lausanne SWITZERLANDBilal Farooq, Swiss Federal Institute of Technology, Lausanne SWITZERLANDAntonin Danalet, Swiss Federal Institute of Technology, Lausanne SWITZERLAND

Inrecentyears,pedestrianflowmodeling,simulation,andoptimizationhasreceivedasurgeinattentionfromthetransportationresearchcommunity.Thefocusofthisattentionismainlyona)solvingthetangibleday-to-dayproblemsofcrowdedpublicspaces,including,transportationhubs,shoppingmalls,etc.andb)optimizingtheevacuationproceduresfromhighoccupancybuildingsandcentersincaseofanunpredictedevent.Thepriorcouldbeofhighinterestforarchitectsfordesignpurposestoseehowpedestriansmoveinbuildingsandalsofortransportengineerstacklingwithsafetyandtransportationfacilitiesintegrationproblemsinbighubs.Thelaterwouldbeofobvioususeforsecuritiesandeventplannersorganizingbigsportmatches,festivals,concerts,etc.

Incomparisontovehicleflowsimulation,pedestrianmovementsaremoreheterogeneous.Pedestrianscanwalkinanydirectionandcanchangetheirtrajectorytowardsalmostanywherethattheydecideto.Unlikevehiclesthataremachineswithknownandpredictablecapabilitiesandwayofmoving,pedestrians‘motiondependsontheirwayofthinking,andinteractingwithothers.Theseareamongthe reasons for which pedestrian modeling is considered to be more complicated and subtler than trafficflow.

Modelsareconsideredassimplifiedrepresentationofthereality.Infact,thegoalistofindmodelswhichareassimplestaspossible,butatthesametimecouldreflectrealisticbehaviors.Generally,pedestrianmodelingisdoneintwodifferentscalesi.e.macroscopicandmicroscopic.Inmicroscopicmodelseachpedestrianisrepresentedseparatelyasanindividualagentandhis/herbehaviorsareexploredindependently.Whileinmacroscopicmodelspedestriansareanalyzedingroupsandcrowdswherethestateofthesystemisgenerallydescribedbymassdensities,flowandaveragevelocity(Schadschneider,Klingsch,Klüpfel,Kretz,Rogsch,andSeyfried,2008).Severalmodelingapproachesinbothscaleshavebeenalreadyputforwardintotheliterature.Forinstance,wecanmentionfluid-dynamicandgaskineticmodelsthataremacroscopicmodelswherepedestriansdynamicstookinspirationfromhydrodynamicsorgas-kinetictheory.Inrecentyearsmoreattentionhasbeenfocusedonmicroscopicmodeling,whereCellularAutomataandSocialforcemodelscanbecited.

Socialforcemodeldescribespedestrianbehaviorthroughso-calledsocialforceswhereinteractionwithenvironmentandotherpeopleisexplainedbyattractiveandrepulsiveforces.ThismodelusesNewton‘sequationtocalculatetheforces.Basicallytheapproachlookslike:

dv/dt=F(t)+fluctuations

WherethefluctuationtermtakesintoaccountrandomvariationsofbehaviorandF(t)gatherssocialforcesnecessarytodescribehumaninteractions.Thelaterisconstitutedofthreeforces.Theseforcesrepresenttheaccelerationtermdepictingthedesiretoachieveacertainvelocitybythepedestrian,therepulsiveeffectevokedbyotherpedestrians,bordersandwallsand,theattractiveforce(HelbingandMolnar,1995).

Socialforcemodelisthemostoperationalpedestrianmodelthatcurrentlyexists,meaningthatitsusedbythemostwell-knownsimulatorsandinthemajorityofpedestrianprojectsandstudies.Onthecontrarythismodelhasneverbeenmicroscopicly(behaviorally)validated.Ingeneral,validationofmicroscopic pedestrian models is performed by comparing aggregate (macroscopic) model parameters (flows,speeds,densities,etc.)oremergingpatterns(dynamiclaneformation,formationofdiagonalstripsincrossingflows)withempiricaldata.Thesocialforcemodelhasbeenvalidatedinthesamemacroscopicwayforselforganizationphenomenai.e.bottleneckoscillations,laneandstripformation.

Indoingso,ithasbeenshownthatthemodelisabletopredictmacroscopicflowconditionswithreasonableaccuracy.Butitisunclearifitonlyprovidesreasonable„average“macroscopicpredictionsoritisabletodescribeindividualwalkingbehavioraccuratelyaswell(P.HoogendoornandDaamen,2007).Insocialforcemodel,forcesarefirstlydefinedbyphysicalconcepts(Newton‘sequation)andthen have been applied to pedestrian behaviors. In a more microscopic and behavioral approach we sticktothecommonhypothesisthatindividualsmakedifferentdecisions,followingahierarchicalscheme:strategical,tacticalandoperational.Destinationsandactivitiesarechosenatstrategicallevel,routechoicesareperformedatthetacticallevelandinstantaneousdecisionswhilewalkingandinteractingwithothersaretakenattheoperationallevel.Ourfocusisonpedestrianwalkingbehavior,naturallyidentifiedbytheoperationallevel.Discretechoicemodelsgodeeperinbehavioralaspectsofpedestriansreactionsratherthanphysicalones.Takingintoaccountdifferentcharacteristicsofpeopleandchoices,thesemodelsdealwithpedestrians‘behaviorbyexploringthewaytheychoosetheiractivity,theirdestination,theirrouteetc.Asanexample,Robin(2011)proposesawalkingpattern(nextstepmodel)wherepedestrianschoosetheirnextstepamong33alternativesinadiscretechoiceframework.Thismodelismicroscopiclyvalidatedwithrealdatawhichprovesthatitcantoalargeextentreflectrealpedestrians‘walkingbehavior.Onthecontrary,discretechoicemodelsarenotasoperational as the social force model.

Obviously,theideathatcomestomindistryingtointegrateadiscretechoicemodelintothesocialforcemodeltotakeadvantageofbothmodels‘strongpointsandtohaveanoperationalmodelwhichisalsovalidatedintermsofmicroscopicbehavior.Thispaperpresentsanoperationalvalidatedmodelbymergingthesephysicalandbehavioralmodels.Themodelisdevelopedandusedinacommercialsimulator(VISSIM)basedonsocialforcemodel.Itisvalidatedwithtworichdatasets.OneisinSVbuildingatEPFL(EcolePolytechniqueFederaledeLausanne)whereatnotabigentrancehallpeople‘swalkingbehaviorsbecomeextremelyinterestingatcertainperiodsofthedaysincetheytakedifferententrancesandexitsdependingonvariousdestinationsthattheycanreachbypassingthishall.AndtheseconddatasetstemfromLausannetrainstationwhichexperiencesinterestingpassengerbehaviorsasahighlyfrequentedmulti-modaltransportationhub.

OpenPedSim: A framework for pedestrian flow analysis

Mohcine Chraibi, Forschungszentrum Jülich GmbH, Jülich GERMANYUlrich Kemloh Wagoum, Forschungszentrum Jülich GmbH, Jülich GERMANYHubert Klüpfel, TraffGo HT GmbH, Flensburg GERMANY

In recent years pedestrian dynamics has gained more importance and a lot of attention due to continuously growing urban population and cities combined with an increase of mass events. In this contextsimulationsarealreadyperformedandcriticaldecisionsaretakenbasedonthesimulationresults.Forthispurposeseveralsoftwarepackageshavebeenpublished.Mostofthesesoftwarearecommercial,ordonotgiveaninsightintheirfunctionality.Whilethismightbeoflittleimportanceforsomeenduser,itisveryimportantforresearcherstoknowexactlywhatthemodelsareperforming.Thisconstitutesasignificantadvantageintheinterpretationoftheresults.

InthispaperweintroduceOpenPedSim(OpenPedestrianSimulation),anopensourceframeworkforperformingpedestriansimulations.Thisframeworkshouldsupportresearchersbythedevelopmentofnew models by providing a suitable environment with appropriate interfaces.

ThemotivationforstartingOPSwasthefollowing:Inmanyuniversitiesandthesisprojects,alotoftimeisspentonsettingupaproperenvironmentforinvestigatingnewideas.Althoughthemainobjectiveis,forinstancethedevelopmentofanewpedestrianmodel,theworkonutilitieslikeaneditorforthegeometryorfortheinputofdata,atoolforvisualizingtheresultsisanenormousentrybarrier.Anothersideissueisthedefinitionofproperinterfaces.Onlyafterthosestepshavebeensuccessfullytaken,the„real“workstarts.Andusuallythisworkhastobedoneasfastaspossible.Mostofthetime,theresultisacodewhichisneitherreusable,normaintainable,norscalable.

WeareherebyintroducingOpenPedSim,anopensourceframeworkforperformingpedestriansimulations,toaddressthoseissues.TheprimarygoalofOpenPedSimistoprovidestudentsandresearcherswithatoolboxthatwillletthemfocusontheirmaintask,i.e.thedevelopment,calibration,andvalidationofnewmodelsormodelfeatures.OPSiscurrentlyfocusingonevacuation,butwillbeextendedtocoverotherareasaswell.Theseare:passengersexchange,commutertrafficinrailwaystations,etc.Finally,OPSalsoprovidessampledatasetsforcalibrationandvalidation.Wearedevelopingsomestandardsandbenchmarkscenariosforevaluationpedestriansimulations.ThistaskiscloselylinkedtoRimeaandtheIMOvalidationcases.

OpenPedSimishostedatSourcefourge.ThedifferentmodulesofOpenPedSimcanbedownloadedfromhttp://sourceforge.net/projects/openpedsim/files/.Therearealsopre-compiledbinariesforthecommonplatforms(Windows/Linux/OSX).Thelatestcodecanbedownloadedviasvnwiththecommandattheurl:\url{https://openpedsim.svn.sourceforge.net/svnroot/openpedsim}.

Assaidabove,amajoraimofOPSistoacceleratestheworkofstudents,researchers,andprogrammers,thatarefocusingonpedestriandynamics.Therefore,OPSprovideswell-designedinterfacesandfile-formatstoreflectthestepsthathavetobecarriedoutwhensimulatingpedestrianflowsandanalyzingthe results.

Usually,apedestrianflowsimulationcomprisesthefollowingsteps:1.Importingthegeometry(eitherfromCADorfromGIS)2.Specifyingthepopulation(syntheticpopulation,includingactivities)3.Specifyingfurtherparameters(e.g.,environment,hazards)4.Performingsimulationruns(includingparametervariation,stochasticvariance,etc.)5.DataAnalysisandInterpretation

ThesinglemodulesofOPSaredesignedtoprovidethecapabilitiesforperformingeachofthesesteps.Atthesametime,thestepsareseparatedfromeachother(„divideandconquer“).Fromatechnicalpointofview,themodulesarethefocus.Thefollowinglistshowstheirapplicationdomain:1.Editor->Geometry2.Editor->Population3.Editor->Environment4.SimulationKernel->Performingsimulationruns5.Visualizer->Visualizationofthesimulatedagentsoff-lineand/oron-line.6.ReportingTools->DataAnalysis

TheOPSframeworkconsistsoflooselycoupledmodules.I.e.,themodulescanbeusedindividually.Theeditorallowsthecreationofnewscenariosforasimulation.Ascenariocomprisesageometry,informationaboutthepedestriansandotherconstraints.Constraintscanbethestateofroomsordoors,blockedforinstance.Thesearetheinputtothemodels.Theresultsofthemodelsarethenvisualizedusingthemoduledesignedforthatpurpose.

Theresultsarethetrajectoriesoftheindividualpedestriansinthesimulation.Butothermodelscanoutputothertypeofresultsaswell,densityforinstance.Inaddition,ameasurementtoolanalysestheresults. Finally a report is generated.

Averyimportantpointhereisthespecificationofthedifferentinterfaces.Aclearandopenspecificationfacilitatestheintegrationofnewcomponents,inthiscaseofnewmodels.Allinterfacesarenotspecifiedyet.

OpenPedSimactuallyconsistsofthefollowingmodules:1.OPSed:Theeditorprovidesgeometriesforcontinuousmodelsaswellasforcellularautomata.2.OPSgcfm:Isanimplementationofthegeneralizedcentrifugalforcemodel3.OPSfast:IsanimplementationoftheF.A.S.T.cellularautomatonmodel4.OPSffca:IsanimplementationoftheFloorFieldCellularAutomaton4.OPStraVisTo:TheVisualizationofthetrajectories.ItmakesuseofQTandtheVisualizationToolKitlibrary(VTK)forrenderingthescenes.5.OPSreport:ReportingToolscontainsmodulesspecifiedintheOPSframeworkandintegratedintheoverallworkflow.Duetolimitedresources,theimplementationofOPSreporthasnotyetstarted.

Atthemomentthesystemislooselycoupled.Afurtherintegrationofthedifferentmodulesmightbeworthwhile.OneideaistoprovideatoolcalledOPScontrol.Thistoolthenmightrunacertainnumber

ofsimulationsbasedonavalidmodelcreatedwithOPSed.Thenextstepwouldbetoidentifythesimulationrunwhichshouldbeinvestigatedfurtherandrunthissimulationwithadetailedoutput,i.e.thetrajectoriesofallpersonsindetail.FinallyOPScontrolwouldinvokeOPSreporttocreateaPDFcontainingasummaryofthesimulationresults.Ofcourse,therearemanymorepossibilities.Anotherideaistouseseveralsimulationkernelsforthesamescenarioandcomparetheresults„automatically“(i.e.byrunningOPScontrolwiththeappropriate(command-line)parameters.

Large-scale multi-modal evacuation analysis with an application to Hamburg

Dirk Durst, Feuerwehr Kerpen, Kerpen GERMANYSven Hebben, TraffGo HT GmbH, Flensburg GERMANYHubert Klüpfel, TraffGo HT GmbH, Flensburg GERMANYGregor Lämmel, Technische Universität Berlin, Berlin GERMANY

Evacuationisonepossibleoptionwhenfacingnaturalorman-maderisks.Theevacuationofabuildingblock,partofacity,orevenawholecityorreagionisafar-reachingmeasure,though.Therefore,itisusuallyalsothelastmeasureandonlytakenwhenasocialcatastropheisimpending.

ThecityofHamburgwashitbyfloodingin1961.Thehomesof50,000peopleweredestroyedandatotalof315personsdied.Thesituationtodayisnotcomparabletothesituationintheearly1960s.Backthen,manybuildingswerestillbarracksandbuiltorrepairedjustafterworldwarII.Thesefewconsiderationsshowthecomplexcontext,inwhichdecisionsaboutevacuationsaremade.Inordertoreducethecomplexityforthedecisionmakerssimulationsforthepredictionofevacuationtimesandpotentialcongestionordelaysareoneoption.Theycanprovideobjectivecriteriaandmaketheconsequencesofcertainalternativesmoreintuitivebyvisualizingthembasedonwell-knownrepresentationsofthecitylikestreetmaps.

Inthispaperwewillpresentresultsonamicroscopicevacuationsimulationcombinedwithdifferentcalculationsfortheevacuationtime.Fourdifferentmodesoftransportaretakenintoaccount:walking,busses,railway,andcars.Indetail:walkingtoshelters,walkingtobusandtrainstations,busshuttleandlocaltrains,andfinallymotorizedindividualtransportbycar.Themajorresultisarangefortheevacuationtime.Itisintheorderofthreetofourhourswhichfitsintotheoveralltime-frameofninehoursforthetotaltimeavailale(i.e.,theavailablesafeevacuationtimeASET).ASETconsistsofthreehoursforpreparationandwarning,fourhoursforevacuation,andtwohoursasbuffer.ASETmustbelargerthantherequiredsafeevacuationtime(RSET),whichisinourcasedeterminedbythesimulations and calculations.

TheAvailableSafeEgressTime(ASET)islimitedfirstofallbytheexternalrisk.Inourcase,theexternalriskisflooding.Tobemorespecific:weassumedaleakageofoneofthedams.HydrologicalandhydrodynamicsimulationsareroutinelyperformedbythecivildefenseauthoritiesinHamburg.Onthatbasis,adetailedplanexists,providingwarningtimesandatimescheduleforalarge-scaleevacuation.Theoverallscheduleis9hours,forobservationandwarningthereare4hoursand1hour,respectively.Therefore,thecurrenttimeframeforindividualpreparationandmovementisfourhours.

PRE-MOVEMENTORPRE-EVACUATIONTIMEWhenassessingevacuationprocesses,manydifferentphasesofinformationgathering,decisionmaking,communication,andfeedbackcollectionhavetobetakenintoaccount.Thesephasesaretosomeextentoverlapping.Ingeneral,theoveralltimecanbedividedintothefollowingparts:EvacuationTime=Detection+Decision+Alarm+IndividualPreparation+Movement

Onemajoruncertaintyistheindividualpreparationtime.Itmightbeinfluencedbythewarningmechanismsandmessages.Foradifferentcasestudy,theIndonesiancityofPadang,upto50%ofthepopulationstatedtheywouldstayinplaceincaseofaTsunamiwarning(Taubenboeck2009).

WARNINGMECHANISMThereisaplethoraofwarningmechanismsplannedforthisevacuationscenario,rangingfromradiomessagesandSMStocanon-shots,sirens,individualphonecalls,etc.Detailsaregivenin(HPA2011,inGerman)andBSI2011(inGerman).

PLACESOFREFUGEThesafeplacesarewithintheevacuationareaaswellasoutside.InHamburg-Wilhelmsburg,placesofrefuge are either higher floors of private buildings or public buildings prepared as evacuation shelters. Theplacesofsuchsheltersareshowninthewarningbrochuredistributedtoallinhabitants(seefigure).Furtherplacesofrefugeareoutsidetheevacuationarerequiringtransportation.Severalmodesoftransportareemployed:motorizedinidividual,busshuttle,andlocaltrains(S-Bahn).

SIMULATIONMODELThispaperpresentsresultsonthesimulationoftheevacuationofHamburgWilhelmsburg.Inthesimulation,25,000agentsweredistributedintheareaoftheElbeisland,whereWilhelmsburgislocated.ThemodelusedisMATSim(www.matsim.org).Thesimulationwaspartofacomprehensiveassessmenttakingintoaccountpublictransport(bylocaltrainsandbusses),foottraffictoshelters,andmotorizedindividualtrafficontheonehand.Ontheotherhand,capacitiesofsheltersandinfrastructure,firstresponders,andinformationdisseminationweretakenintoaccount.Thetrafficsimulationmodulewasusedtodeterminetheoverallevacuationtimeformotorizedindividualtraffic.Sincethemovementtosheltersandthetransportationbypublictransport(localtrainsandbusses)isindependentthereof,themaximumofthetimesforthedifferentevacuationmodescanbeconsideredthe overall evacuation time.

FIRSTRESULTSInthescenarioanalyzedanadvancewarningof3hourswasassumed.Thistimeisalsousedforthepreparation of the emergency and rescue service.Formotorizedindividualtraffictheoverallevacuationtimeislessthan90minutes.However,thetransportationbypublictransporttakesaboutfourhours,whichusesuptheavailablesafeevacuationtimegivenlocalauthorities.TheservicetimesforthepublictransportevacuationwasbasedonanevacuationexerciseperformedinCologne.Inordertoimprovethesituationsanincreaseoftheavailabletransportinfrastructure(i.e.numberofbusses/trains)oramoreeffectiveusageoftheexistingtransportinfrastructureisneeded.Thiswouldmeantoincreasethenumberofpersonsperbusortrain.Ithastobenotedthatthenumberofpersonsperbuswaslessthan20%asundernormalconditions.Thisisduetotheextensiveamountofluggagecarriedbytheevacuees.

FURTHERSTEPSTheresultspresentedhereareembeddedintoalargercontext,whichistheGRIPSresearchproject.MoredetailsaboutGRIPSaregivenbyWalencialet.al.(intheseproceedings).

Poster exhibition

Unsorted list

Daichi Yanagisawa, College of Science, Ibaraki University, Ibaraki JAPAN Starting-wave and optimal density in a queue

Xuan Xu, China Academy of Safety Science and Technology, Beijing CHINA Optimal exit choice algorithm for pedestrian evacuation dynamics

Dongyoun Shin, Eidgenössische Technische Hochschule, Zürich SWITZERLAND Activity classification and user interface design for a crowdsourcing urban simulation platform using mobile devices

Stefan Nowak, Universität zu Köln, Köln GERMANY A cellular automaton model for lane formation in bidirectional pedestrian flow

Antonin Danalet, Swiss Federal Institute of Technology, Lausanne SWITZERLAND Pedestrian map matching using WiFi traces

Hubert Klüpfel, TraffGo HT GmbH, Flensburg GERMANY Specific requirements for the egress simulation of large scale outdoor events

Khalidur Rahman, Universiti Sains Malaysia, Georgetown MALAYSIA Analysis of pedestrian walking speed in a developing country: A factorial design study

Starting-wave and optimal density in a queue

Daichi Yanagisawa, College of Science, Ibaraki University, Ibaraki JAPANAkiyasu Tomoeda, Meiji Univeristy / JST, CREST, Tokyo JAPANKatsuhiro Nishinari, University of Tokyo, Tokyo JAPANTakashi Imamura, University of Tokyo, Tokyo JAPAN

Variouskindsofself-drivenmany-particles(SDP)systems,suchaspedestriandynamics,vehiculartrafficandtrafficphenomenainbiologyhaveattractedagreatdealofattentioninawiderangeoffieldsduring the last few decades. Most of these complex systems are interesting not only from the point ofviewofnaturalsciencesforfundamentalunderstandingofhownatureworksbutalsofromthepoints of view of applied sciences and engineering for the potential practical use of the results of the investigations.Especially,theinterdisciplinaryinvestigationsforthedynamicsofjammingphenomenainSDPsystems,so-calledJamology,havebeenprogressedbydevelopingsophisticatedmathematicalmodelconsideredasasystemofinteractingparticlesdrivenfar-fromequilibrium.Thesecontributionstoanalyzethemechanismofjammingformationtellthatoneofthemostimportantfactorstocausethejammingphenomenaisthesensitivity,whichindicatesthetimedelayofreactionofpedestriansordriverstothestimulus.Asanexample,ifthereactionsofdriversareextremelysensitive,theycanavoidthetrafficjambyadjustingtheirbehaviorimmediatelytotheirfrontcar’smovement.Thereactiontimeofpedestriansissimilarlyimportanttowardsmoothmovementofcrowd.Moreover,wewouldliketopointoutthatthewaveofsuccessivereactioninaqueue,so-calledstarting-wave,playsasignificantroleforthewaitingtimeofqueuingsystemofpedestriansandvehicles,sincequick-startinwalkingaccomplishesthemoresmoothmovementofcrowds.Inthiscontribution,firstofallweinvestigate the relation between the propagation speed of pedestrians’ reaction and their density by usingmathematicalmodelbasedonthestochasticcellularautomata.Then,theoptimaldensitytominimizethetraveltimeoflastpedestriansinaqueuetoreachtheheadpositionoftheinitialqueueisinvestigatedundertakingintoaccountthepropagationtimeofstarting-wave.Finallyweverifytheseresults obtained from mathematical model by performing the experiments of pedestrians.

Ourmathematicalmodelisbuiltonthestochasticcellularautomata,whichrecentlyprevailstomodelthestochastictransportincomplexsystems.LetusimaginethatthepassageispartitionedintoLidenticalcellsthateachcellcanaccommodateatmostoneparticle(pedestrian)atatime.Notethat,inthefollowing,wereferto“particle”asarepresentationofapedestrianinamodeland“pedestrian”asapersonitself.Thelengthofeachcellcorrespondsto0.5metersbyconsideringthereasonablevolumeexclusioneffectofpedestrians.Moreover,atotalnumberofNindicatestheparticleswhichareplacedatequaldistanceHcell.Theupdaterulesofourcellularautomatonmodelareasfollows:firstofall,onlytheparticleatfrontofaqueuemovesforward.Thenthefollowingparticleonlythesecondparticleinthequeuecanmoveforward.Afterthesecondparticlemovesforward,thenextparticlecanstarttomoveinsequence.Theserulesofpedestrians’walkingareappliedinparalleltoallparticles.Notethat,inourmodel,inordertoinvestigatethepropagationspeedofsuccessivereactions,allofthefollowingparticlescannotmoveforwardbeforethestarting-wavereachestothem.Therefore,unliketheusualstochasticcellularautomatonmodelssuchasASEP,ZRP,inourmodel,onlyifthenextcellisempty

andpredecessorhadalreadymoved,followingparticlescanmoveforwardwithprobabilityp(h)whichdependsontheirheadwaydistanceh.Thishoppingprobabilityofparticlesp(h),whichindicatesthevelocityofparticles,isgiveninanalogywiththeideaofOptimalVelocity(OV)function,whichisoftenintroducedintothemathematicalmodelforvehiculartrafficasadesiredvelocityofdriversdependingonheadwaydistance.Thisfunctionismotivatedbythecommonexpectationthatdrivershavetheirdesiredvelocitytodothecomfortabledriving.Wehavemeasuredthepropagationspeedofstarting-wavewhichisderivedfromthelengthofinitialqueuedividedbytheelapsedtimeundereachgiveninitialdensitywhichisdecidedbytheinitialvalueofparticlesonthecells,thatis,N/L.Theresultsobtained from mathematical model show the power law in the relation between propagation speed of starting-waveandtheinitialdensityofpedestrians.

Ifthereisanoptimaldensitytominimizethedelaytogetoutfromaqueueorcrowd,thecontrolofdensityispossibletoreducethewasteofwaitingtime.Moreover,ifpedestriansstandinalinewithlargeheadway(lowdensity),thestarting-wavepropagatesfastunderthelongqueue.Whereas,iftheystandinalinewithsmallheadway(highdensity),thestarting-wavepropagatesslowlyundertheshortqueue.Whichsituationdecreasesthewasteofwaitingtime?Theoptimalinitialdistributionforaqueueisinvestigatedherebyconsideringthefundamentalrelationcharacterizedbythepowerlaw.Theresultsobtainedfrommeanfieldtheorybasedonthemathematicalmodelshowthattheoptimaldensitydoesexistatadensity,whichdependsontheirwalkingvelocity.

Wehavefoundthattheseresults,thatis,fundamentalrelationcharacterizedbythepowerlawandtheexistenceofoptimaldensity,areverifiedbyperformingtheexperimentsofpedestrians.

Inthiscontribution,wehaveinvestigatedthepropagationspeedofpedestrians’reactioninrelaxationprocessofaqueueso-calledstarting-wave.Thefasterthestarting-wavepropagates,themoresmoothlythecrowdmoves.Wehaverevealedtheexistenceofoptimaldensity,wherethetraveltimeoflastpedestrianstoreachthestartlinefortheinitialqueueisminimizedbybothanalyticalcalculationsandexperiments.Thisoptimaldensityinevitablyplaysasignificantroletodesignnotonlytheinitialqueueofpedestriansbutalsothetrafficproblems.

Optimal exit choice algorithm for pedestrian evacuation dynamics

Xuan Xu, China Academy of Safety Science and Technology, Beijing CHINAMaohua Zhong, China Academy of Safety Science and Technology, Beijing CHINACongling Shi, China Academy of Safety Science and Technology, Beijing CHINA

Inthemicroscopicevacuationmodel,thedesireddirectionofthepedestrianisconsistentwiththeminimum-distancedirection.However,theminimum-distanceroutemaynotsameastheminimum-timeroute.Theevacuationtimeismadeupwiththemovementtimetotheexitandwaitingtimeattheexit.Themovementtimeisdeterminedbytheexitlocation,whilethewaitingtimeisdeterminedbypedestriandistributionandexitwidth.Themicroscopicmodelusuallysolvestheexitchoicebasedonthedistance,andotherconstraintsarerarelytakenintoaccount.However,insomecases,theshortestdistancedoesnotnecessarilyguaranteetheminimumevacuationtime.Forexample,whenthereisahighoccupantdensityaroundthenearestexit,thepedestrianwillspendalotofwaitingtimeifhechoosesthenearestexit.Instead,ifhechangeshisoptiontoselectafartherexit,hemayleavemorequicklyifthereisalowoccupantdensityaroundtheexit.Moreover,supposetherearetwooptionalexits,oneofwhichhasashorterdistanceandsmallerwidth,theotherhasalongerdistanceandlargerexitwidth.Choosingthenearerexitratherthanthefartheronemayspendmoretimebecausetheexitwidthistoosmalltoletthecrowdpassquickly.Thekeyofsolvingthisissueistohelpthepedestrianknowhowlonghewillwaitattheexit.Thewaitingtimeisrelatedtothepedestriandistributionandexitwidth.Thus,theindividualevacuationtimeisnotonlyrelatedtothedistancebutalsothepedestriandistributionandexitwidth.Whenchoosingtheexit,notonlytheexitposition(evacuationdistance)shouldbeconsideredinestimatingtheevacuationtime,butalsothepedestriandistributionand exit width.

Inthispaper,theoptimalexitchoicealgorithmispresentedbasedonthemicroscopicmodel.Thealgorithm aims at the shortest individual evacuation time which is calculated in iterative method. Thefactorssuchaspedestriandistribution,exitpositionandexitwidthareallinvolvedinthetimeevaluationfunctioninthealgorithm.Takingaroomwithmultipleexitsasanexample,thecellularautomatamodelisappliedtosimulatetheevacuationprocess,andtheevacuationtimewithoptimizedandnon-optimizedexitchoicearecomparedandanalyzed.

Tobenoticed,forthereasonthatduringtheevacuationprocess,nervousandherdingmightinfluencetheexitchosenbehavior,theoccupantsmightnotchangetheirdecisionaslongastheexittheychoseisavailable.Asaconsequence,inthepresentarticlewedonotintendtointroducedynamicalmethodtodeterminetheexitoptionoftheoccupants.Whatismore,frequentlychangingsignalofexitsignscould induce potential crowd disaster when occupants follow these signals and change their moving directionsfromtimetotime.Asaresult,theoptimalexitselectionmethodisastatisticalgorithm.Theexitchoiceforeachpedestrianisvisuallydisplayedbytheexitselectionzoningmap.Thezonesoccupiedbypedestrianswithdifferentcolorsareassociatedwithdifferentexits.Eachzonecoveredwith the same color corresponds to one exit.

Theresultsshowthatundertheinfluenceoftheoptimizedexitchoice,somepeoplenolongerchoosethenearestexitbuttheexitwithlargerwidthorwithloweroccupantdensityaround.Thealgorithmsolvestheproblemsofunevenandinadequateutilizationoftheexitintheevacuation.

Astheexitchoicechanged,theevacuationtimethrougheachexitchangedaccordingly.Theoverallevacuationtime,namelythemaximumvalue,isthetimewhenallthepedestriansleavetheroom.Itisfoundthattheoverallevacuationtimedropsfromthe850sto564sfortheoptimizedexitchoice,whichindicatesthattheevacuationefficiencyincreasesby33.64%.

Thestandarddeviationoftheevacuationtimecanreflecttheuniformityoftheexitutilization.Thehigherthevalueofthestandarddeviationis,thegreaterthedifferenceofexitutilizationis.Theresultsshowthatthestandarddeviationoftheevacuationtimefortheoriginalexitchoiceis195s,whilethevaluefortheoptimizedexitchoiceis145s.Itindicatesthatadoptingtheoptimizedexitchoicerealizedthediversionofthecrowdflowduringtheevacuation,andthewaitingtimewasdistributedapproximately evenly.

Idletimeofanexitisdefinedasthedifferencebetweentheoverallevacuationtimeandtheevacuationtimeoftheexit.Itisfoundthatcomparedtotheoriginalexitchoice,theidletimeforallexitswithoptimizedexitchoicedecreasesgreatly.Fortheoriginalchoice,theaverageidletimeis520sandthemaximumidletimeis781s.Whentakingoptimizedchoice,theaverageidletimereducesto226sandthemaximumidletime454s.Theaverageidletimedropsby56.54%,whichdemonstratesthattakingoptimizedexitchoiceeffectivelyreducesthevacancyrateoftheexitandenhancestheutilizationgreatly.

Theoptimalalgorithmgivestheexitselectionzonemapwhichcanprovidereferenceforplanningevacuation.Asoneevacueecannotgetglobalinformationofallevacueesintheroom,theymaychoosethenearestexitbutneedtowaitforalongtime.Thus,itisnecessarytoguideevacueestothemostproperexitsfromtheglobalperspective.Thedividinglinebetweendifferentexitselectionzonesisthekeytoallocateevacuees.Incaseofemergency,theexecutorshouldassignsomeguidersnearthedividinglinetoconductevacueestoanoptimalexitaccordingtothezonemap.Infact,theexitselectionzonemapfromtheresultsprovidestheplacementofguidersandtheguidingdirectiontothe designer.

Activity classification and user interface design for a crowdsourcing urban simulation platform using mobile devices

Dongyoun Shin, Eidgenössische Technische Hochschule, Zürich SWITZERLANDStefan Müller, Eidgenössische Technische Hochschule, Zürich SWITZERLANDSungah Kim, Sungkyunkwan University, Seoul SOUTH KOREAGerhard Schmitt, Eidgenössische Technische Hochschule, Zürich SWITZERLAND

INTRODUCTIONANDRESEARCHQUESTIONSTheresearchareaofurbansimulationmethodshasgrownnotablyinrecentdecades.Mostoftheresearchtopicsthatconcernurbansimulationhaveconcentratedondefiningthecomplexitiesofurbanenvironmentswithcertainrulesandalgorithms.However,citiesaregettingmorecomplexandchangestothemarebeingmadeatgreaterspeed[2].Therefore,currenturbansimulationmodelingapproaches based on rules and protocols are still struggling to reduce the gap between the virtual simulationenvironmentandtherealcities,sincethebehaviorofcitizensisfrequentlyunpredictableand continuously adapting.

Inthiscontext,researchisnecessarytodevelopmorefundamentalsimulationmethodsthatcanhandlethesecomplexitiesandchanges,leadingtonewdesigndecisionsupportsystems[3].Therefore,thisresearchwasmotivatedwiththefollowingquestions:Whatistheoriginofthecomplexitiesandtransformationsoftheurbanenvironment?Howcanweapproachtheorigintodealwiththeurbancomplexities and transformations?

Toanswerthesequestions,wehypothesizethatthediversehumanbehavioraretheoriginoftheissuesthat result from all of the complexities and changes of the cities.

CROWDSOURCINGURBANSIMULATIONPLATFORMANDKEYWORDSInthispaper,weintroducetheideaofacrowdsourcingurbansimulationplatformusingsmartphones[1,4].Suchdevelopmentrequiresresearchtobeconductedinnumerousdisciplines:socialsensing,urbansustainability,mobilenetwork,behaviorpatternanalysis,andsocialnetworkservices.Itisbasedonthedetectionandclassificationofactivitypatterns(currentlymainlyfocusingontrafficinformation),andonstate-of-the-artinteractiveuserinterfacesinordertoletuserutilizetheapplicationeasily.Onthebasisofthesecutting-edgetechnologies,thisresearchaimsatthedesignandimplementation of a practical participatory urban sustainability simulation platform.

OBJETIVESAsaresult,weproposeaparticipatorysimulationenvironmentthatfeedssensedhumanbehaviorintoanurbansimulator,andtherebyincludesurbancomplexitiesanddynamicssimultaneously.Thisresearch pursues to collect urban behavior data through the smartphone and suggests possible user benefitsfortheirdatasharing.

Thereforethisresearchhasfollowingobjectives:

-EstablishsensingmethodsformassdatacollectionusingsmartphonesThefirststepofthisresearchwastosetupthebasisformobilecrowdsourcing.Thisincludesdefiningwhatkindsofdatashouldbecollected,howthedataneedstobetransformedtobeusefulforurbansimulation,andhowinformationcanbefedbacktotheuser.

-Classifytheuser’sactivity,focusingontransportationmodeWehavedevelopedaprototypemobilephoneapplicationthatimplementsanoveltransportationmodedetectionalgorithm.Themobilephoneapplicationrunsinthebackgroundandcontinuouslycollectsdatafromthebuilt-inaccelerationandnetworklocationsensors.Thecollecteddataisanalyzedbythe transportation mode detection algorithm and automatically partitioned into activity segments. Akeyobservationofourworkisthatwalkingactivitycanberobustlydetectedinthedatastream.Thereforeitisusedasaseparatorforpartitioningthedatastreamintootheractivitysegments.Eachvehicleactivitysegmentisthensub-classifiedaccordingtothetypeoftakenvehicle.OurapproachyieldshighaccuracyatalowsamplingintervalanddoesnotrequireGPSdata.Therefore,devicepowerconsumptionisminimized.

-DesignandimplementaneffectiveuserinterfaceforcontributingandsharingInordertomakecrowdsourceddatacollectioneffective,massparticipationisessential.Thereforeitisessentialtoattractusersandtokeeptheminterestedinanapplication.Thus,wewillintroducethepossiblealternatives;showhowtoeffectivelydeliverandvisualizetheusers’information,howtogivebenefitstousersinreturnfortheircontributions,andsuggestwhatkindofsocialinteractionscanbeimplemented to induce more participation.

FUNDAMENTALGOALSANDCONCLUSIONSOurgoalistoenablepeopletoshareurbaninformationatanytimethroughthecrowdsourcingurbansimulationplatform[5].Theinformationwillbereturnedtothecitizenstosupporttheirsustainability-awarelife.Thesimulationplatformalsogivesachancenotonlytocompareeachother’slevelsofsustainability,butalsotogiveself-satisfactionthroughanaltruisticcontributionforasustainablefuture.Thus,peopleshallutilizethesimulatorinordertopredicttheirindividualorcities’futuresustainability.Meanwhile,theuserdatawillbecollectedanddeliveredtothecentralserverinordertoanalyzetheurbansustainability.

Wepresentthemethodscollectingurbandatausingsmartphoneandactivityclassificationfromtheaccelerationandlocationdata,andfurther,theuserinterfaceanduserbenefitaresuggestedtogiveamotivateforvoluntaryparticipation.Consiquently,wecanmeasuretheurbansustainabilitybasedonarealhumaninteraction,andcompareindividualsaswellascities.Thewholeprocessofthisresearchis presented as a new paradigm of an urban simulator that reflects the urban complexities and the inconstant human mind changes.

REFERENCES

[1] Crowdsourcing: The term „crowdsourcing“ is a neologistic portmanteau of „crowd“ and „outsourcing,“ first coined by Jeff Howe in a June 2006 Wired magazine article „The Rise of Crowdsourcing“ (Howe 2006). Howe explains that because technological advances have allowed for cheap consumer electronics, the gap between professionals and amateurs has been diminished. Companies are then able to take advantage of the talent of the public, and Howe states that „It’s not outsourcing; it’s crowdsourcing.“ (Wikipedia)

[2] Aiberti, M. (1996). „Measuring urban sustainability.“ Environ Impact Assess Rev 16: 381-424.

[3] Bell, S. and S. Morse (2008). Sustainability indicators: measuring the immeasurable?, Earthscan/James & James.

[4] Brabham, D. (2009). „Crowdsourcing the public participation process for planning projects.“ Planning Theory 8(3): 242.

[5] Halatsch, J., A. Kunze, et al. (2010). „ETH Future Cities Simulation Platform.“ The Design of Material, Organism, and Minds: 95-108.

A cellular automaton model for lane formation in bidirectional pedestrian flow

Stefan Nowak, Universität zu Köln, Köln GERMANYAndreas Schadschneider, Universität zu Köln, Köln GERMANY

Inapedestriancrowdithappensoftenthatnotallpeoplehavethesamedestination.Evenopposingdesiredwalkingdirectionsinpedestriandynamicsaretheruleratherthantheexception.Whenthisisthecase,oneusuallyobservesthatthepedestrianssegregateintodistinctlaneswhereeachlanecontainsonlypeoplewithequalwalkingdirection.Thislaneformationprocessisawellknownemergentbehaviorinbidirectionalpedestrianflow.Butitcanalsobeobservedinsomephysicalsystems,e.g.,inabinarymixtureofoppositelychargedcolloidalparticlesthataredrivenbyanexternalfield[1].Thosesystemswithcounterflowandlaneformationareveryinterestingfromaphysicalpointofview:Theyaretypicalexamplesforself-organizationprocessesandspontaneoussymmetrybreaking.Additionally,onecanoftenobservenon-equilibriumphasetransitionsandhysteresisbehavior.Therefore,theywereanalyzedquitefrequentlyinphysics,butinpedestriandynamicsitisnotmuchknownquantitativelyaboutthisphenomenon.

Formanymodelstheoccurrenceoflanesisusedasatestfortherealism,butthestatementsareonlythatlanesarepresentinthesystem.Amoredetaileddescriptionoftheformationprocessandthepropertiesoflanesareveryrare.Furthermore,simulationsofpedestriancounterflowareoftenperformedtoanalyzeaso-called``jammingtransition‘‘[2],i.e.,thefocusisnotonthelanesbutonthejamsinthesystem.Butthereisnoempiricalevidencethatsuchatransitionshouldoccurinreality.Ingeneral,comparisonswithempiricaldataconcerningcounterflowareextremelyrare.Thisisrelatedtothe fact that up to now also the empirical situation is not very satisfying.

Hereweanalyzelaneformationinatwodimensionalcellularautomatonforpedestriandynamics.Themodelisbasedonthefloorfieldcellularautomatonmodel[3].Itdividesthespaceintodiscretecellsandrepresentsthepedestriansbyparticlesonthecells.Thebasicinteractioncomesfromtheconstraintthateachcellcanbeoccupiedbyatmostoneparticle,buttheimportantfeatureinregardtolaneformationisthe„dynamicfloorfield‘‘.Itcanbeinterpretedasavirtualtracewhichindicatesthecellswhichwereoccupiedbyaparticleinthenearbypast.Thistracehasanattractiveeffecttoparticlesofthesamedesiredwalkingdirection.Theeffectisthattheparticlestendtofolloweachother,whichissufficienttocreatelanes.Butatlargedensitiesthisbasicmodelbecomesunrealistic:Thedynamicsistoo much influenced by collisions between particles and for larger densities the system always evolves in a complete jam.

Asaconsequenceonehastointroduceadditionalmechanismsinordertoavoidjamsmoreeffectively.Oneofthemistogivethevirtualpedestrianstheabilitytolookahead.Whenpedestriansareinacounterflow,theyusuallytrytoavoidcollisionswithotherpedestriansbyestimatingtheprospectiverouteofpedestrianwithoppositewalkingdirections.Thisbehaviorcanbeimitatedinthismodelby

introducinganadditional`ànticipationfloorfield‘‘.Itindicatesthespacewhichismostlikelytobeoccupiedbyparticlesinthenearbyfutureandhasarepulsiveeffecttoparticlesoftheoppositedesiredwalkingdirection.Anotherextensionofthebasicmodeltakesintoaccountthatpedestriansareabletowalkpasteachotherevenifthereisveryfewspace,i.e.,iftheavailablewidthislessthantwocells(=0.8m).Theyareforinstanceabletoturnsidewayssuchthatthecontactareabecomessmaller.Thisisincluded in the model as the ability of two particles to swap their position under certain conditions.

Toquantifythelanestructure,wemakeuseofalaningorderparameter,whichhasbeenusedalreadybyRexandLoewentodetectlanesinasystemofoppositelychargedparticles[4].Thiscanhelptoestimateveryeasilyunderwhichconditions(modelparameters,environment,etc.)laneformationispossible and gives new insights into the phenomenon.

Furthermore,wecomparetheresultsofthemodelwithempiricaldatafromacounterflowexperiment[5]whichwasperformedrecentlywithintheproject„Hermes“.Datafortheflow-densityrelationshipwerecollectedwithhighaccuracy.Wecalibratethemodelparameterssuchthatthedataarereproduced.Inaddition,thereweretwoversionsoftheexperiment.Inthefirstversion,theparticipants get no instruction about which of two possible exits (left ore right) they have to choose. Oneobservesacompletesegregationofthetwoopposingstreamsandlanesareformedimmediatelyafterthebeginningoftheexperimentandstaystablethewholetime.Inasecondversion,theyweretoldtochooseanexitaccordingtoanumbertheygotbeforetheexperiment:Oddnumbersexittotheleft,evennumberstotheright.Onestillobserverslaneformationinthisversion,butthelanesarenowunstableandtheirnumberincreases.Thisbehaviorcanalsobereproducedbythemodelwiththeparameters obtained from the fundamental diagram.

REFERENCES

[1] J. Dzubiella, G. P. Hoffmann and H. Loewen , Lane formation in colloidal mixtures driven by an external field, Phys. Rev. E, Vol. 65 (021402), 2002

[2] Takashi Nagatani, Freezing transition in bi-directional CA model for facing pedestrian traffic, Physics Letters A, Volume 373, Issue 33, 2009

[3] C. Burstedde, K. Klauck, A. Schadschneider and J. Zittartz, Simulation of pedestrian dynamics using a two-dimensional cellular automaton, Physica A, Volume 295, 2001

[4] M. Rex and H. Loewen, Lane formation in oppositely charged colloids driven by an electric field: Chaining and two-dimensional crystallization, Physical Review E, Vol. 75 (051402), 2007.

[5] J. Zhang, W. Klingsch, A. Schadschneider, A. Seyfried, Ordering in bidirectional pedestrian flows and its influence on the fundamental diagram, arXiv:1107.5246v1, 2011.

Pedestrian map matching using WiFi traces

Antonin Danalet, Swiss Federal Institute of Technology, Lausanne SWITZERLANDMichel Bierlaire, Swiss Federal Institute of Technology, Lausanne SWITZERLANDBilal Farooq, Swiss Federal Institute of Technology, Lausanne SWITZERLAND

Inthispaper,wewillpresentthedatacollectionmethodologyandamapmatchingalgorithmadaptedforpedestriansandpassiveWiFitraces.

Gatheringdataaboutpedestrianlocalizationandtrackingisdifficult,inparticularindoorandonalargescale.Camerasallowonlyasmallareatobecovered.GPSdevicesdonotworkindoorandface problems with distribution and access to the data of a large enough population. For a whole transportationhuboracommercialcenter,itisoftennottechnicallyappropriateandsociallyaccepted.Thesedataareimportantforroutechoicemodeling,descriptionofcongestion,efficientdesignofnewfacilities,andtravelguidanceandinformationsystems.

WeproposetouseWiFitracestotrackpedestrianpaths.Thisdatacollectionworksindoor,isnotinvasivefortheuser,andworksproperlyforlargefieldsofobservation.WiFitracesallowforthelocalizationofmobiledevicesandthusoftheirusers,butarescarceandfuzzy.SinceWiFitracesarenotnecessarilyrelatedtothepedestriannetwork,weneedtodevelopamapmatchingmethodologyadaptedtothedataandtothepedestriancontext.OurapproachoffersawaytodefinepedestrianpathsindoorfromWiFitraces.WewilltestitontheEcolePolytechniqueFédéraledeLausanne(EPFL)campus.

EPFLCampushosts11’800persons,including7600studentsand4200employees.Differenttypesoffacilitiesaredistributedon55hectares.Likemanypublicandprivatebuildings,EPFLisextensivelycoveredbyWiFiwith789accesspoints.Itallowustotrackpedestriansonthewholecampus.Forourexperiments,weusetheroutingtoolforpedestriansavailableatEPFL(http://plan.epfl.ch)asapedestriannetworkrepresentation.Unliketraditionalrouteguidancemapservices,italsotakesintoaccount the floors and stairs in the buildings.

WecollectlogsofcampusmembersfromtheWiFiaccesspoints.Wealreadyhaveaccesstoonedatasetandweplanaseconddatacollection.Thefirstdatasetgivesthelocationoftheaccesspointtowhichauserisconnected.Intheseconddatacollection,wecollaboratewithCiscoinordertolocalizemorepreciselytheusers.Weusethesignalsreceivedbyallaccesspointsaroundtheuserandnotjusttheonetheuserisconnectedto.Someoftheselogsarestaticdevices,othersarequasi-mobileusedonlyinafewspecificplacesoncampus(likelaptopsinclassroomsandatthelibrary),buttherearealsosomemobiledevicessuchassmartphones.Itwillallowustotrackpedestriansonthecampus.Theprecisionisquitelow(10metersradiusin90%ofcases,5meters10%ofremainingcases,dependingonthecontext),butitworksinthebuildings,wheremostofthetripstakeplacewheretheGPSdevicesdonotwork.Thismethodofcollectingdatahasseveraladvantagesoverdevice-centeredprocesseslikethe

onesusingsmartphones,sinceitdoesn’tneedtohaveformalacceptancefromtheusers.Thisprocessistransparentfortheusers,anditalsoworksfortheusersthatarenotauthorizedtoconnecttothesystem.IfanexternalvisitorcomestoEPFL,evenifhedoesn’thavethecredentialrequiredtoconnecttotheWiFinetwork,hissmartphonewillannounceitsMACaddress,whichisauniqueidentifierandthereforeallowsustotrackeachdeviceTheMACaddressisanonymizedbyEPFLITDepartmentinorder to guarantee privacy.

ThepoorqualityandthescarcityofWiFilocalizationprecludestheuseoftraditionalmapmatchingmethods.WeplantoadaptaprobabilisticmapmatchingmethodrecentlydevelopedinourlabforSmartphoneGPSdata(M.Bierlaire,J.ChenandJ.P.Newman,ModelingRouteChoiceBehaviorFromSmartphoneGPSdata,TechnicalReport101016,TransportandMobilityLaboratory,EPFL,2010).Wewillgenerateasetofpotentialtruepathsandassociatealikelihoodwitheachofthem.Thismethodproposesaframeworkbasedonmeasurementandstructuralequationsadaptedforfuzzyandscattereddata.

WewillmatchasetofpathswithWiFitraces.Foreveryindividual,thealgorithmconsiderseachWiFitracechronologically.Ateachiteration,itgeneratesasetofpathcandidatestopologicallyextendingthepreviousone.Thisextensionisnotmadewithonearcbutwithmany.Allthesearcsareinthedomainofdatarelevance(aconceptintroducedbyBierlaireandFrejingerin2008inanarticleinTransportationResearchPartC)oftheWiFitrace,dependingontheprecisionofthemeasurement.Foreacharcinthisdomainandeachendofpathintheprevioussetofpathcandidates,thealgorithmconnects them with the shortest path.

Thenweupdatethelikelihoodofeachnewpossiblepath.Forthispurpose,wewilluseaprobabilisticmeasurement model. It computes the probability that an individual with a mobile device could have generatedasequenceofWiFilogswhilefollowingagivenpath.Itisbasedonastructuralmodelandameasurementmodel,whichcapturesthemovementsofpedestriansandthetracesoftheWiFiaccesspointsrespectively.BothwillneedtobeadaptedforpedestriandynamicsandWiFitraces.

ThepossibleapplicationsofthismethodologywithWiFitracesarepotentiallylarge.Itwillbeusedoncampus,butitcouldhelptounderstandpedestrianbehaviorwithalowcostindatacollectionfortransportationhubs,commercialcentersandmaybeevenforcitycenters(e.g.shoppingstreetsegments)orformassevents,iftheyarecoveredbyWiFi.

Specific requirements for the egress simulation of large scale outdoor events

Hubert Klüpfel, TraffGo HT GmbH, Flensburg GERMANYSven Hebben, TraffGo HT GmbH, Flensburg GERMANY

INTRODUCTIONEgresssimulationforlargescaleoutdooreventsisachallengeforegresssimu-lation.Thisholdswithrespecttomethodandinterpretationandassessmentofresults.Theaimofthiscontributionistoconvincethereaderthategresssimula-tioncanmakeavaluablecontributiontoincreasesafety(andcomfortofvisitors.Tothisend,thelimitsoftheapplicationofegresssimulationareaddressedaswellastheconnectiontofiresafetyengineeringandsafetyconcepts.Withoutbeingembeddedinsuchalargercontext,thesimulationresultsareuseless.

SimulationprogramsaretoolslikeCNCmachines.Neithertheaimofitsappli-cationnortheinterpretationandassessmentofresultsarecreatedinthesimu-lation.Bothaspectshavetobeprovidedfromtheoutside(bytheuser,theeventmanager,theauthorities,etc.).

If an evacuation analysis is able to provide a substantial contribution to the planning of an event can be decided by the persons who are responsible for the planning and safety. If an event is safe must be decidedbywhomwhoisre-sponsiblefortheassessmentoftheresultsortheissuingofapermitofconduc-tance.Thisissuewillbeaddressedindetailinsection5attheend.

LARGESCALEEVENTSLargescaleeventsareachallengeforsafetyandsecuritystaff.InGermany,aspecialsafetyconceptisrequired.AftertheLoveparadedisasterinDuisburg,theministryoftheInteriorofthestateofNorth-RhineWestphaliaissuedtofol-lowingdecree:[7]:“InthecontextoftheeventsattheLoveParade2010inDuisburg,thefollowingclarificationconcerningtheprocessofpermissionisgiven:Forthestategov-ernment,thefollowinghasforemostpriority:• Thesafetyconceptforlargescaleeventsmustberenderedthoroughlyandinconsentwithall stakeholders.• Theauthoritywhichisresponsibleforthepermissionmustensuretheparticipationofallother relevantauthoritiesinatransparentandcompre-hensiveprocess.

AsafetyconeptisrequiredaccordingtoSBauVONRW§43forvenueswithmorethan5000visitorspaces.

Itmightbenecessaryforsmallervenues,too,dependingontheinfrastructure,therelationofthesizeofthevenuetothesizeofthetown.

VALUESANDAIMESLegalregulationsareforthesafetyofpeople(respectivelycitizens)andtheirrights.Thoseare:lifeandhealth,propertyandpublicsafety.Thebuildingregu-lationsexplicitlystate“publicsafetyandorder,specificallylife,healthandnatu-ralfoundationsoflife”.Tosavethosegoods,specificaimsmustbedefined.Thoseaimsarefortheirprotection:• Thespreadofafiremustbehindered(foracertaintime)• Theextinctionofafiremustnotcausedamagewhichisoutofproportion

Theevacuationofabuildingorvenueisameasuretoprotecthealthandsafetyofinhabitantsandvisitors.Theemergencyegressisordertodothis.Andthis,ofcourse,alsoholdsfortheevacuationprocess itself: the egress paths must be dimensioned such that they allow for safe and orderly egress in anemergencysituation.Butthisisnotenoughforthesafetyofanevent.

Therearemanyfurtherrequirements:TheGermanAssociationofProfessionalFireBrigades(AGBF)hasissuedachecklistcomprisingninechapters[1]1. Introduction2. CrisisManagement/CrisisTeam/CrisisStaff3. Processforsafetyrelevantissues4. Evacuation/EmergencyEgress5. Massiveoccurenceofinjuries(MANV)6. ConceptforSecurityStaff7. ConceptforFirstResponders8. ConceptforFireSafetyStaff(SWD)9. Appendicesa. Checklistsb. GuidelineforFireSafetyaccordingto14096(ifapplicable)c. PlansforFireBrigadesaccordingtoDIN14095(ifapplicable)d. Flucht-undRettungsplänenachDIN4844

TheChecklistisfreelyavalaible[1](inGerman).Thedocumentexplicitlystatestheaimtocontributetofairnessbetweeneventandvenuemanagers.Tothisendclear,concise,andreliableinformationabouttherequirementsforsafetyconceptsisprovided.Therequirementsmightcauseconsiderableeffortandshouldthereforebepredictableandclearlydefined.Thesafetyconceptmustbeagreeduponbythepolice,firebrigade,ambulanceservice,etc.Thisalsore-quiresacleardefinitionofitsscopeandcontents.

LOVEPARADETheLoveparadedisasterhappenedinDuisburginJune2010.Wewillanalyzethecausesandreasonsfor the disaster from a systemic perspective and derive “soft” criteria for detecting pitfalls in the planning process.

CONCLUSIONInthissectionwewillsummarizetheargumentspresentedintheprevioussec-tions.Thisleadstorecommendationsconcerningthesafetyoflargescaleevents.Thefocusisontheplanningphase,theassessmentofsimulationre-sults,thesevenmythsofmasspsychology,andthemethodofequivalenceanalysis.

REFERENCES

[1] Arbeitsgemeinschaft der Leiter der Berufsfeuerwehren in der Bundesrepublik Deutschland (AGBF): Sicherheitskonzepte für Versammlungsstätten, 28.03.2008. http://www.agbf.de/f08_loveparade-i01.htm

[2] Vabeg Eventsafety GmbH: Referenzprojekt „Reichsstraßenfest Donauwörth“

[4] RiMEA – Richtlinie für Mikroskopische Enfluchtungsanalysen. Version 2.2.1, 8. Juni 2009. www.rimea.de

[6] TraffGo HT GmbH. Informationen zum Umfang des Auftrags der Lopavent GmbH für die Loveparade 2010 in Duisburg (Angebot, Entfluchtungsanalyse und Anlagen) www.traffgo-ht.com/temp/temp-ub1-del122011/ub1.html

[7] Erlass des Innenministers NRW vom 11.08.2010, Az. 71/38.05.01. Wortlaut und Erläuterungen unter http://www.im.nrw.de/sch/819.htm

[8] Landtag NRW, Ausschussprotokoll Apr 15/7 vom 04.08.2010. http://www.landtag.nrw.de/www/www.landtag.nrw.de/portal/WWW/dokumentenarchiv/Dokument/MMA15-7.pdf

[11] Mamrot D.: Lovaparade Duisburg 2010 – Wi(e)der die normale Katastrophe. Ingenieurforum 3/2010, Seiten 38 bis 41. www.vfdb.de/downloads

[12] Mamrot D. und Blätte H.J.: Loveparade Duisburg 2010 – Vorschlag zu einer neuen Planungsmethode für Veranstaltungen mit Katastrophenpotential. Vfdb-Zeitschrift 4/2010, Seiten 93 bis 95.

[13] Oberhagemann D.: Analyse der Besucherzahlen und der Ereignisse auf der Rampe zum Veranstal-tungsgelände während der Loveparade 2010 in Duisburg. (30.07.2010). Vereinigung zur Förderung des deutschen Brandschutzes (vfdb), http://www.vfdb.de/download/AnalyseLoveparade2010.pdf

[14] Nanda Wijermanns: Understanding Crowd Behavior. Dissertation, Universität Groningen, 2011. http://www.evacmod.net/?q=node/2556

Analysis of pedestrian walking speed in a developing country: A factorial design study

Khalidur Rahman, Universiti Sains Malaysia, Georgetown MALAYSIANoraida Abdul Ghani, Universiti Sains Malaysia, Georgetown MALAYSIAAnton Abdul Basah Kamil, Universiti Sains Malaysia, Georgetown MALAYSIAAdli Mustafa, Universiti Sains Malaysia, Georgetown MALAYSIA

ABSTRACTThedirectuseofforeigndesigncodesandunavailabilityofwell-recognizedlocalparametersforpedestrianfacilitieshasbeenaconcernindevelopingcountries.Inthispaper,astudyonpedestrianmovementsonsidewalksinthecapitalcityDhaka,Bangladeshhasbeendonetoidentifythecontributingfactorsonpedestrianmeanspeeds.Basicdataonwalkingspeedsandfactorsthatmightaffectthespeedswerecollectedfrom1,440pedestriansbyaphotographicprocedureofvideorecording.FactorialDesignwithMixedLevelswasused.Resultsshowthatwalkingspeedsaregreatlyaffectedbypedestrianage,genderandwidth/densityofthefacility.BangladeshipedestriansareslowerthanthoseofWesterncountries,butarefasteroralikecomparedtosomeAsiancounterparts.SuchfindingdoesnotvalidatethesustainabilityoftheadoptionofforeigndesignandparametersforpedestrianfacilitiesinBangladesh.Theresultsofthisanalysiscanbeusedasaguidelinefordevelopingdesign codes for local pedestrian facilities.

INTRODUCTIONToreduceenvironmentalpollution,pedestrianizationhasbecomeanintegralpartofsustainablemodernurbandesign.Toachieveso,pedestrianfacilitiesshouldbeplannedandbasedontheconcreteinformationonusercharacteristics,travellingpatternsandobjectivesofpedestrianflow.Thesmoothmovementofpedestriansisaffectedbyanumberoffactorsincludinggenderofpedestrians,andlocationofwalkingfacility[1],widthofthewalkingfacility[2],baggagecarryingcapacity[3].PedestriansarethemostvulnerablegroupinDhakaCitywhenitcomestoroadaccidents,astheyconstitute51percentofthevictimsoftrafficfatalities[4].ThisapprehensionmotivatedthecurrentstudyforstatisticalanalysisofpedestrianindividualspeedvariationonsidewalksinDhaka,Bangladeshandtoexaminetheinfluenceofsomefactorssupposedtoaffectthespeeds.

METHODOLOGYSomeparticularlocationswereselectedfordatacollection.Observedmovementofpedestrianswerebi-directionalwithnoentryfromorexittootherwalkwaysandpedestrianswerewithdifferenttripobjectives.Aphotographicprocedurewasusedtocollectthebasicrelevantdataofpedestrianmovements.Dataonpedestrianmovementsatselectedlocationswererecordedduringpeakandoff-peakperiodsonthreetypicalweekdaysinSeptemberof2011underclearanddryweathercondition.

Combinationofthelevelsoftheindependentvariablesage,gender,carryingbaggageandlocationmade36cells.Theanalysisofvariance(ANOVA)forFactorialDesignwithMixedLevelswasusedtoanalysetheinfluenceofthechosenvariablesonthewalkingspeedsofpedestrians.Incaseswherethenullhypothesiswasrejected,theleastsignificantdifference(LSD)methodwasusedforpairwise

comparisonoflevel-means.Alltestsweredoneat1or5percentlevelofsignificance.SPSS(StatisticalPackageforSocialScience)wasusedtoanalysethecollecteddata

MAINRESULTSTable1:Listofpedestrianmeanspeedindifferentcountries

Continent Country Author Pedestrianmeanspeed(m/sec)Asia SaudiArabia [5] 1.08 India [6] 1.20 China [7] 1.20 Bangladesh 1.20 Thailand [1] 1.22 Singapore [8] 1.23Europe England(UK) [9] 1.31NorthAmerica USA [10] 1.47 Canada [11] 1.40

Pedestrianmeanwalkingspeedsfromdifferentcountries(whichalsoincludetheresultofpresentstudy)arelistedinTable1.Surprisingly,themeanspeedofBangladeshipedestrians,1.20m/sec,isequivalenttothemeanspeedsofIndianandChinesepedestrians.BangladeshisadjacenttoIndiaandlinkedwithChinabyIndia.Thisfindingmightsuggestthatsocioeconomicconcordamongthepedestrianshasagreatinfluenceonwalkingbehaviors.ItalsobearsoutthatBangladeshipedestriansareslowerthanthoseofWesterncountries,butarefasteroralikecomparedtosomeAsiancounterparts.SuchfindingsdonotvalidatethesustainabilityoftheadoptionofforeigndesignandparametersforthepedestrianfacilitiesinBangladesh.HenceanappropriatepedestriandesignstandardisverynecessaryforAsiacontinent,particularlyforSouthAsia.

FromtheANOVA,itwasfoundthatallmainfactors-age,gender,location(i.e.widthofthelocation)andbaggage-andtwointeractionsgender-baggageandgender-locationsignificantlyaffectthewalkingspeedsofpedestriansat1percent.

ACKNOWLEDGEMENTBothIPSandtheResearchUniversity(RU)GrantScheme,[Acct.No.:1001/PJJAUH/811097],UniversitiSainsMalaysiaareacknowledgedbythefirstauthoronlyandallauthorsforUSMfellowshipandtheuseofresourcesthroughtheconductofthisstudy,respectively.

REFERENCES

[1] Tanaboriboon, Y. and J. Guyano, Analysis of pedestrian movements in Bangkok. Transportation Research Record, 1991(1294).

[2] Mitchell, D.H. and J. MacGregor Smith, Topological network design of pedestrian networks. Transportation Research Part B: Methodological, 2001. 35(2): p. 107-135.

[3] Fruin, J., Designing for Pedestrians: A Level-of-Service Concept. New York Metropolitan Association of Urban Designers and Environmental Planners. Highway Research Record, 1971.

[4] Rahman, M., Respecting the Rights of Pedestrians, in Dhaka Courier2010: Dhaka.

[5] Koushki, P.A., Walking characteristics in Central Riyadh, Saudi Arabia. Journal of transportation engineering, 1988. 114(6): p. 735-744.

[6] Victor, D. Pedestrian traffic management in Indian cities. in International 5th conference on transport research. 1989. Yokohama, Japan.

[7] Lam, W.H.K., J.F. Morrall, and H. Ho, Pedestrian flow characteristics in Hong Kong. Transportation Research Record, 1995(1487): p. 56-62.

[8] Tanaboriboon, Y., S.S. Hwa, and C.H. Chor, Pedestrian characteristics study in Singapore. Journal of transportation engineering, 1986. 112: p. 229.

[9] Older, S., Movement of pedestrians on footways in shopping streets1968: Traffic engineering & control.

[10] Hoel, L.A., Pedestrian Travel Rates in Central Business Districts. Traffic engineering, 1968. 38: p. 10-13.

[11] Morrall, J.F., L. Ratnayake, and P. Seneviratne, Comparison of central business district pedestrian characteristics in Canada and Sri Lanka. Transportation Research Record, 1991(1294).

Poster exhibition

Unsorted list

Andrea Gorrini, University of Milano-Bicocca, Milano ITALY An innovative scenario for pedestrian data collection: the observation of an admission test at the University of Milano-Bicocca

Kensuke Yasufuku, Osaka University, Osaka JAPAN Scalable evacuation simulation and visualization using GPU computing

Wassim Abu Abed, Leibniz Universität Hannover, Hannover GERMANY Cognition-oriented simulation of pedestrian dynamics: Perceptually driven social force model

Winnie Daamen, Delft University of Technology, Delft NETHERLANDS Interaction behaviour between individual pedestrians

Bani Anvari, Imperial College London, London UNITED KINGDOM Shared space simulation based on social forces and distance potential field

Julien Cividini, University Paris-Sud, Paris FRANCE Frozen shuffle update in simple geometries: A first step to simulate pedestrians

Burkhard Forell, Gesellschaft für Anlagen und Reaktorsicherheit, Köln GERMANY Comparison of evacuation simulation models

An innovative scenario for pedestrian data collection: the observation of an admission test at the University of Milano-Bicocca

Andrea Gorrini, University of Milano-Bicocca, Milano ITALYMizar Luca Federici, Crowdyxity s.r.l., Milano ITALYLorenza Manenti, University of Milano-Bicocca, Milano ITALYGiuseppe Vizzari, University of Milano-Bicocca, Milano ITALY

Inthecontextofpedestrianandevacuationdynamics,oneofthemostrelevanttopicisrelatedtothedifficultyindatacollectionaboutthisphenomenon[1,2]:thesedatarepresentusefulinformationtocharacterizeeventsandsituations,buttheyarealsonecessarytovalidatemodellingandsimulationstudiesofpedestrianandcrowddynamics.Fromamethodologicalpointofview,experimentsandobservationsarethemoresuitabletechniquestocollectdata:inthisworkwechosetoperformanobservationofarealevent,theadmissiontestoftheUniversityofMilano-Bicocca,whichtookplaceinSeptember1st,2011.

Thissurveywasaimedatgatheringempiricaldatarelatedtopedestrianandgroupdynamicsinasituationofmedium-highdensity,intheveinofpreviousobservationsandmodellingproposals[3].Thesurveyconsistsofaquantitativedatacollection,basedonapeoplecountingactivitysupportedbyvideoanalysisoftheevent,withthefinalaimofdevelopingandvalidatingbehaviouralmodelsandwhat-ifscenariosimulations.Theanalysisisfocusedontworelevantaspectsoftheoveralldynamics:compositionandbehaviourofgroupsofpedestriansandproxemicbehaviour[4],chosenasananalyticalindicatorofcrowdbehaviouraldynamics,thankstoitsabilitytomodelinterandintra-grouprelationshipsinhigh-densitysituationsbasedontheregulationofspatialdistance.Insituationsofhigh-mediumdensity,theproxemicbehaviourofwalkinggroupsischaracterizedbytypicalpatternsofspatialdistribution(line-abreast,V-like,river-likepatterns)[5],dependingonthepsychologicalbondingandtheneedofpreservingthepossibilityof(alsononverbal)interactionsamongmembers,theneedtoavoidphysicalcontactwithothers,andthefeaturesofthegroup’smembers(e.g.,gender,age,andsoon)[6].

Inthiscontext,theaimofthecasestudywastoobserveindividualandgrouppedestrianbehaviourbeforetheopeningtime,andduringtheentryprocessinthevenuesoftheadmissiontesttotheFacultyofPsychologytotheUniversityofMilano-Bicocca,whichwasattendedbyabouttwothousandpeople(2094students,including437males(29%),and1657females(79%)).Thedatacollectionphasewasperformedbymeansofunobtrusiveobservationandheadcountingactivityfromdifferentplannedlocations(thatweredeterminedafterapreliminaryon-sitevisitinordertocaptureflowsfromallaccesspoints),anditthewasfocusedon:generalizedcountingofpedestrianflow,groupsize,groupspatialarrangement(degreeofalignmentandgroupcohesion),groupwalkingspeed,andformationofqueues.Toensurethevalidityoftheresearch,inaddictiontothepresenceofredundantobserversandsupervisors,video-recordinginstrumentswerealsoemployedduringtheprocessofdatacollection.Inthecontextofpedestriandatacollection,observationsareusuallyperformedintransportplaces,suchastrainstationsandairports[7],andcommercialplaces,suchasshoppingmallsandcommercialwalkways[3].Forthisreason,theanalysedscenarioisaninnovativecasestudy.Moreover,evenifthe

admissiontestisanindividualtask,excludinggroupparticipation,thedataanalysisunderlinesthatmorethan65%oftheobservedincomingflowwascomposedofgroups,and,inparticular,ofcouples(77%),triples(19%),andgroupsoffourmembers(4%).Anon-goingdataanalysis,relatedtoaselectedportionofthepedestrianflow,isfocusedontherelationshipamong:walkwaylevelofservice,socialdensity,groupsizeandshape,gender,andwalkingspeed.

Startingfromthiswork,thepaperwillgiveacompletedescriptionofthesurveyactivity,startingfromaexhaustivetheoreticaldiscussionabouttherelationshipamongcrowd,pedestriangroupsandproxemics,andasystematicdescriptionofthecasestudy:thescenarioanalysis,themethodologicalapproach,thedatacollected,andtheperformedanalysis.Thepresentedcasestudycanbeusefultosupportdevelopmentsinthemodellingcrowdandpedestriandynamics,bymeansofsimulationtools:thedataanalysiswillbeexploitedtoimplementseveralwhat-ifscenariosrelatedtotheobservedphenomenon.ThankstothecollaborationwiththeauthorityoftheUniversityofMilano-Bicocca,theresultsachievedcanbeausefulstartingpointtodesignalternativestrategiesrelatedtoamoreefficientmanagementofpeoplewhoattendeveryyeartheadmissiontest.Theresultsofthisworkwillsupportthedevelopmentofpoliciesandguidelinesforthemanagementofattendees,there-organizationofthephysical layout of the environment (by means of the use of barriers or signposting) in order to reduce queuingtimes.

REFERENCES

[1] Peacock, R. D., Kuligowski, E. D., & Averill, J. D. (2011). Pedestrian and Evacuation Dynamics. Springer Verlag.

[2] Resilience, U. K. (2009). Understanding Crowd Behaviours. Cabinet Office, London.

[3] Moussaïd, M., Perozo, N., Garnier, S., Helbing, D., & Theraulaz, G. (2010). The walking behaviour of pedestrian social groups and its impact on crowd dynamics. PLoS One, 5(4), e10047.

[4] Hall, E. T. (1966). The Hidden Dimension. Doubleday, New York.

[5] Karamouzas, I., & Overmars, M. (2010). Simulating the local behaviour of small pedestrian groups. Proceedings of the 17th ACM Symposium on Virtual Reality Software and Technology, pp. 183-190.

[6] Costa, M. (2010). Interpersonal Distances in Group Walking. Journal of Nonverbal Behavior, 34(1), pp. 15-26.

[7] Schultz, M., Schulz, C., & Fricke, H. (2010). Passenger Dynamics at Airport Terminal Environment. Pedestrian and Evacuation Dynamics 2008, pp. 381-396.

Scalable evacuation simulation and visualization using GPU computing

Kensuke Yasufuku, Osaka University, Osaka JAPAN

Real-timesimulationtoolsareeffectiveforinteractivelyanalyzingevacuationbehaviors.However,computingandvisualizingmassiveagent-basedcrowdsinreal-timeisacomputationallyintensivetaskbecausemostalgorithmsforcalculatingtheinteractionsamongalltheagents.Inpreviouswork,researchershavereducedthiscomplexitybyemployingspecialdatastructures,suchasmeshmodelsornetworkmodels.Additionally,researchershavesignificantlyincreasedcomputationalperformancebyadaptingexistingCPU-orientedalgorithmstoparallelprocessingarchitectures.Apromisingnewparallelarchitectureusescommoditygraphicsprocessingunits(GPUs)havingmanycores,offeringtheperformancebenefitsofparallelprocessingatalowcost.ThiswastheadventofthemovementcalledGPUcomputing.Usingthistechnology,itisnoweasiertodesignscalableagent-basedsimulationsthatcanbeexecutedonGPUs.Inthispaper,wewilldescribeournovelapproachtoreal-timevisualizationofcrowdflowusingascalableevacuationsimulationthatissuitableforexecutiononGPUs.

Theevacuationsimulationusesanagent-basedsystem.Namely,anevacueeismodeledasanagent.Themodelingofeachagent’sbehaviorconsistsof(1)theroutechoicemodel:settingadestinationandcalculatingtheroute;and(2)thecrowdwalkingmodel:approachingthedestinationandavoidingcollisionswithotheragentsandobstacles.Inourresearch,weimplementedashortestroutechoicemodelandthesocialforcemodel[Helbingetal,2000].

Conventionalmethodsimplementagent-basedmodelingusingCPU-orientedalgorithms.ThesocialforcemodelthatweimplementedpreviouslywasalsocalculatedontheCPU,withtheagentsrenderedontheGPUonlyaftertheyhaveupdatedtheirposition.Howeverinrecentyears,theuseofGPUsforgeneralpurposecomputinghasbecomeanewareaofresearch.Inournewsimulation,thesocialforcemodeloperationswereexecutedonaGPUusingCUDAtechnology,whichisaparallel-processingarchitectureforGPUswithmanycores.Specifically,onethreadisallocatedperagenttocalculatethesocialforcesonthatagent.AlthoughCPUscanonlyconcurrentlyexecuteonethreadpercore,CUDAcanruntensofthousandsofthreadssimultaneously.ThustheGPU’scomputationalpowerissufficientforupdatingthebehaviorofscalableagent-basedcrowdsinreal-time.

Toevaluatetheperformanceofoursimulationmethod,werantwoimplementationofthesimulation.ThefirstwasexecutedonaGPU,andthesecondwasexecutedonaCPU.Thesamealgorithmanddatastructureswereusedinbothsimulations.ThesetestswereconductedonanIntelCorei79302.80GHzCPUandanNVIDIAGeForceGTX460GPU.Thesetestswereexecutedforeachimplementationtype,varyingonlythenumberofagents,whichrangedbetween1000and10,000.ItwasfoundthattheGPUversionhadbetterscalabilitythantheCPUversion.Forexample,at10,000agents,theGPUversionwasapproximatelyseventimesfasterthantheCPUversion.Moreover,itcanbeclearlyobservedthat

theperformanceoftheGPUversionwassufficienttosustaininteractiveframeratesforrenderingcomplexmodelsofagentsandbuildings.Suchinteractivevisualizationiseffectiveforinteractivelyanalyzingevacuationbehaviorsinthetrial-and-errorstageofcreatinganevacuationplan.

Beforeapplyingtheevacuationsimulationtoalargespace,weverifythecrowdwalkingcomponentof the social force model using a much simpler space by calculating the relationship between crowd walkingspeedandcrowddensity.Asaresult,anincreaseinthecrowddensitycausesadecreaseinwalkingspeed.Thistendencyconformstoanempiricalmodel.

Asacasestudyofscalableevacuationsimulation,wecreatedanevacuationscenarioforalargeundergroundshoppingmallinOsaka,Japan.Thisundergroundshoppingmallhas1,200retailstoresandrestaurants,aswellasasubwayandintercityrailstation.Thusthisundergroundareahasoneofthemost complicated space compositions.

Asaresult,ittook5minutestoevacuateallthe25,000peopleinthelargeundergroundshoppingmallaccordingtotheevacuationsimulation.Howeverapproximately80%ofevacueesescapedwithinonly1minute.Therefore,itislikelythatthereareareaswherethenumberofstaircasesisinsufficient.Althoughsomemightclaimthatthereisnopracticaluseinsimulatingasituationinwhichevacueesareuniformlydistributedandallevacueesescapethroughoneoftheneareststaircases.However,wethinkthissimulationcanserveasaguideforanalyzingtheevacuationofthisarea.

Inthispaper,wehavedescribedascalableevacuationsimulationthatcanbeexecutedonGPUswhilevisualizingcrowdflowinalargeundergroundshoppingmallinteractively.Oursimulationhasdemonstratedthatthesocialforcemodel,whichisusedinthissimulationtomodelthewalkingbehavioroftheagents,isareasonableapproximationofrealcrowdwalkingbehaviorintermsofwalkingspeedandcrowddensity.Moreover,wedemonstratedthattheGPU-basedimplementationwascapableofsupportingupto25,000agentsataninteractiveframerateusingcurrentgraphicshardwareandCUDAtechnology.Thereforewehaveshownthatthissimulationhaspotentialasaneffectivetoolfor designing and evaluating large urban environments.

REFERENCE

Helbing, D., Farkas, I. and Vicsek, T., „Simulating Dynamical Features of Escape Panic“, Nature (2000), Vol.407, pp.487-490

Cognition-oriented simulation of pedestrian dynamics: Perceptually driven social force model

Wassim Abu Abed, Leibniz Universität Hannover, Hannover GERMANYVolker Berkhahn, Leibniz Universität Hannover, Hannover GERMANY

Simulationmodelsofpedestriandynamicshaverecentlybecomeimportanttoolsofplanninganddesigninmanyareasofarchitectureandcivilengineering.Theyaidthearchitectorengineerinhisorherattempttocreatesecureandcomfortablepedestrianfacilitiesandenvironments.Theincreasingimportance of pedestrian simulations emphasises the need of accurate and reliable simulation models. Forthesimulationofpedestriandynamicsawidevarietyofdifferentmodelsexitsrepresentingthesedynamicsanalogoustophysicalsystems.Despitethefactthatmostofthemodelscurrentlyinusereproduce many of the available empirical observations of pedestrian dynamics relatively good on a qualitativelevel,reliableresultsonthequantitivelevelarestillmissing.Thephysicaloriginandnatureof the observation strategies and the modelling approaches are responsible for the incomplete and insufficientrepresentationoftheobservedsystem.Hence,thecomplicatedbehaviourofhumanbeingscannot be realistically reproduced by just applying a physical scheme of stimulus and reaction.

Acognition-orientedapproachtomodellingpedestrianmotioncantakeintoaccounttheflexibilityofhumanbehaviour.Heresignificantresearchisneededforarealisticsimulationofpedestrianbehaviourthatisnotrigidlycoupledtothestimulioftheoutsideworld,butratheraresultoftheinternalcognitiveprocesses.Thereproductionofthemissingflexibilityofthehumancognitionandthe resulting behaviour is an important development for compensating the limited analogy between pedestriandynamicsandphysicalprocessesinthemathematicalmodelssofarinuse.Aparadigmchangeshouldbeundertaken,inwhichapedestrianisnotmerelyanobservedmovingphysicalobjectanymore,butratheranobservedintelligenthumanbeing.Thefindingsandmethodsofcognitivescienceprovideasoundandwell-groundedbasistocopewiththischallenge.Themainpurposeofcognitivescienceistoexplaintheinternalprocessesthattakeplaceinsideahumanbeingafterreceivingastimulusandbeforetakingtheoriginatedortriggeredreaction.Asystematicadaptationofthemethodsandfindingsofcognitivesciencecandrasticallyimprovethemodellingmethodologyofpedestrian dynamics and will open the door for further multidisciplinary collaboration with a wide rangeofdisciplines,especiallycognitivepsychology,architectureandcivilengineering,andcomputerscience.

Thispaperconsidersmodellingthespatialperceptionofpedestriansasoneofthekeycognitiveactivities.Sightandhearingaretheprimarysensesresponsibleforthespatialperceptionofthesurroundings.Externalstimuliusuallyfindtheirwaytothespatialcognitivesystemofahumanbeingthroughthismental,physiologicalcapability.Hence,spatialperceptionisanimportantfoundationof the behaviour of pedestrians and has the primary function in facilitating interactions between the individualanditsenvironment.Thespatialrepresentationoftheenvironment,sofarinuseinthecurrenttheoreticalphysicalsimulationmodelsofpedestriandynamics,takesabird’s-eyeviewofthe

surroundings.Theexactpositions,sizeandmotionisrepresentedinapreciseeuclideangeometry.Thisrepresentationsatisfiesthepurposeofdescribingamotionmodel,butitdoesnothelptofacilitatethe modelling of a perceptually guided interaction between individuals and their environment. Therefore,aspatialrepresentationfromapedestrianstandpointorperspectiveisneeded.Eachindividual pedestrian interprets the spatial information he perceives in terms of his own subjective frameofreference.Incontrasttothepreciserepresentationusedinmotionmodels,thedesiredspatialperceptualrepresentationshouldtakeintoaccountthesubjectiveinterpretationofthesurroundingsand the associated state of mind or feeling emphasising internal processes not resulting directly from the stimulus input. Modelling the spatial perception on the basis of such an egocentric spatial representationwillbeastartingpointofcognition-orientedsimulationsofpedestriandynamics.

In this paper a mathematical model of the subjective human spatial perception will be presented. Motivatedbythefactthathumancognitiveactivitiesareapproximateratherthanpreciseinnature,themodelisbasedonthetheoryoffuzzysets.Thispedestrian-centreddynamicrepresentationofthesurroundingisthenintegratedinamodifiedversionofthewell-knownsocialforcemodeltoresultinaperceptuallydrivenmotionmodelofpedestrians.Bridgingthegapbetweentheflexibilityofhumanrepresentations and the precision and clarity of physical or computerised representations is a major advantageofthisapproach.Severalimprovementsinsimulatingtheindividualandcollectivebehaviourofpedestriansareexpected.Avalidationalongsideasensitivityanalysisoftheresultingperceptuallydrivensocialforcemodelwillalsobeprovided.Theperformanceofthemodelisevaluated.Itscapabilities,limitationsandpossiblefurtherimprovementswillbediscussed.Thevalidationprocessismakinguseofqualitativeandquantitivecomparisonswithempiricaldatareportedintheliterature.Thepaper will end up with a discussion of the achieved results and a presentation of the drawn conclusions. Thefocusoffutureinvestigations,suchas,thepossibilityofmodellingtheperceptionrelatedeffectsofstresssituationsassociatedwiththeso-called“Panic”behaviour,andthepossibilityofsimulatingpedestriandynamicsindarkorsmoke-filledenvironmentsinmorerealisticway,willbehighlightedamongotherfurtherfutureresearchwork.

Interaction behaviour between individual pedestrians

Winnie Daamen, Delft University of Technology, Delft NETHERLANDSSerge Hoogendoorn, Delft University of Technology, Delft NETHERLANDSDirk Versluis, Delft University of Technology, Delft NETHERLANDS

Recently,publicplacessuchasshoppingmallsandpublictransportterminalshavebecomemoreandmorecrowdedduetoincreasedpopulationsizeandmobility.Theeaseofuseoftheseplacesforpedestriansdependstoalargeextentonthepedestriandensityinpeakhours.Sincespaceinthesedenselyoccupiedareasishighlyvalued,theuseofeverysquaremeterhastobeevaluatedcarefully.

Plannersanddesignersofpedestrianinfrastructurehavealargeneedforaccurateevaluationtools.Verypromisingtoolstofulfilthisneedaremicroscopicpedestriansimulationmodels.Thesecanprovideanobjectiveassessmentofapedestrianfacilityandquantitativelypredictcriticalareasandbottlenecksinthefacility’sdesign.

Animportantfeatureofthesemicroscopicmodelsistheabilitytorepresentindividualwalkingbehaviourinarealisticway.Anessentialelementinindividualwalkingbehaviouristheresponsetothebehaviourofotherpedestriansthatareencounteredinthewalkingarea.Thisisreferredtoasinteractionbehaviour.Especiallytheinteractionsbetweenindividualpedestriansappeartobepoorlysimulatedonthemicroscopiclevel.Thispaperdescribestheseinteractionsinmoredetailusingempirical data.

Toidentifyandquantifypedestrianinteractionmovementswehavecollecteddetaileddatainlaboratory experiments. In laboratory experiments the external conditions and other factors that influenceinteractionbehaviourcanbecontrolledandtheeffectoftheseseparatefactorscanbededucedmoreeasily.Withasmartexperimentaldesigntheamountofinformationthatcanbeobtainedfromalaboratoryexperimentcanbemaximized.Basedonaliteraturestudy,wehaveselectedthefollowingexperimentalvariables:age,bodysize,gender,freespeed,travelpurpose,manoeuvrability,numberofpedestriansandwalkingarrangements.First,observationsontheintendedpathhavebeenperformed,wherepedestriansarenothinderedbyotherpedestrians.Then,pedestriantrajectoriesfortheinteractionsituationsaremeasured.Walkingoutsideoftheintendedpathinunhinderedwalkingimpliesthatinteractionmovementsareperformed.Interactionmovementscanbequantifiedbymeasuringlateralandlongitudinalevasionfromthemeantrajectoryandmeanspeedgraphinunhinderedwalking.Thefollowingexperimentshavebeenperformed(N=Normal,H=Hurry,O=nOreducedmanoeuvrability,R=Reducedmanoeuvrability): Agroupoftwelveparticipants(sixmenandsixwomen)participatedintheexperiments,inordertohavesufficientrepetitionsofsimilarencounters.Thelengthofthewalkingstretchis20meters,whereinteractionisexpectedtohappenhalfwaythislength,i.e.at10metersfromtheinitialstartingposition.

Usingthepedestriantrajectories,verydetailedbehaviouralanalyseshaveperformed,bothfortheunhinderedandfortheinteractionsituations.Forunhinderedwalking,wehavestudiedwalkingspeed,steplengthandstepfrequency.Withthisknowledge,wearebetterabletoidentifywhetherthebehaviourduringinteractionsisduetonaturaldeviation,orwhetheritshouldbeattributedtotheinteraction.Forinteractionbehaviour,wehavelookedatlateralandlongitudinalinteractionmovements,andwehavecomparedforthevariousinteractionsituationsandpedestriancharacteristicspassingsideandextentofevasion(bothinlateralandlongitudinaldirection).Inthepaper,wewillpresenttheextensiveanalyses,herewelimitourselvestoourfindings.

Ithasbeenfoundthatindividualpedestriansperformmovementsthatarerelatedtointeractionin88%ofalloccasionswhentheymeetanotherpedestrian.Theseinteractionmovementsconsistoflateraland/orlongitudinalevasivemanoeuvrestoavoidacollision.Thesemovementscanbeinterpretedassome gallantry towards other pedestrians.

Thesidewherepedestrianspasseachotherisfoundtobedependentonthedirectionofapproach.Forcrossingsituationsthesideofapproachhasnoinfluenceonthepassingside.However,iftheangleofapproachincreasesandthesituationcomesclosertobidirectional,pedestriansprefertopasseachother on the right hand side. In the bidirectional situation pedestrians strongly prefer passing on the righthandside.Itisalsofoundthatwalkinginahurryincreasestheprobabilityofpassinginfrontof another pedestrian in crossing situations. Meeting a small group of two pedestrians increases the probabilityofpassingattheback.

Thepassingsideandthedirectionofapproach(i.e.bidirectionalorcrossing)mainlydeterminethedirectionandextentofevasionfromtheindividualmeanwalkingpathandfromtheindividualmeanwalkingspeed.Pedestriansseemtopreferlargerlateralevasioninbidirectionalsituationsandlargerlongitudinalevasionincrossingsituations.Moreover,menlaterallyevademorethanwomenandhurriedpedestrianslaterallyevademorethannormallywalkingpedestrians.Itseemsthathurriedpedestrianareeithergranted‘rightofway’inanearlystageoftheinteractionprocessorthattheytaketheinitiativeininteractionthemselves.Finally,theextentofevasionislargerwhensmallgroupsareencountered.

Shared space simulation based on social forces and distance potential field

Bani Anvari, Imperial College London, London UNITED KINGDOMWinnie Daamen, Delft University of Technology, Delft NETHERLANDSSerge Hoogendoorn, Delft University of Technology, Delft NETHERLANDSVictor Knoop, Delft University of Technology, Delft NETHERLANDSMichael G.H. Bell, Imperial College London, London UNITED KINGDOM

Thedesignofpublicspaceincitiesprovidesopportunitiesforleisure,socialinteractionandphysicalactivitieswhichinfluencethetravelpattern,roadcongestionandtrafficdelays.Urbandesignmovestowards changing the way streets function by reducing the dominance of vehicles. In this new urban designprinciple,avarietyofstandardisedmechanisms,controlsystemsandmarkingsarereducedcreatingastrongrelationshipbetweenthestreetanditssurroundings.Thus,sharedspaceusersneedtorelyoneye-contact,negotiatetheirrightofwaywithothertrafficparticipants,anddrivemorecarefullyinordertocorrectlyreadsituations.Thereisaneedforaquantitativeassessmentofsharedspaceprinciples in order to identify the conditions under which sharing street space with pedestrians is a feasiblealternativetotraditionalstreetdesigns.Asapartofthisaim,amathematicalmodelisproposedtodescribethemainbehavioursofsharedspaceusersbasedonsocialforces.Inaddition,aDistancePotentialFieldusingthefloodfillmethodisgeneratedforfindingtheshortestpathtowardsthetargetdestination.Realdatafromthreesharedspaceenvironments:Brighton(UnitedKingdom),Haren(TheNetherlands)andBohmte(Germany)areusedtosupportthebehaviouralassumptionsofthesimulation and will be used to calibrate model parameters in the future.

TheproposedmathematicalmodelisbasedontheSocialForceModel(SFM),whichexplainstheaccelerationofanobject(likeapedestrian)inatwo-dimensionalspaceastheresolutionofforcesexertedbyneighbouringobjects(suchasotherpedestrians,cars,andfixedobstacles)andthetarget(destination).TheSFMresemblesacarfollowingmodelextendedtotwodimensions,whichhasthepotentialtodescribegapacceptancebehaviouraswell.AnadvantageoftheSFMisthatitsparametersareassociatedwithmeaningfulquantitiesthatcanbemeasured.Inaddition,theSFMcanbemodifiedtoallownewobjectsand,differentbehavioursoractionstobeincluded.Thisisaccomplishedbychangingthevalueofparameters,suchasthedesireddirectionofmovement.Theresultingmodelallowsmulti-directionalflowstobesimulated.ThesefactorsallcontributetothechoiceoftheSFMasthe mathematical basis for the simulation and evaluation of shared space schemes.

Sincepedestrianmovementandcartrafficcoexistwithinsharedspaceenvironments,thebasicSFMforpedestriansismodifiedbytheadditionofnewobjectsrepresentingvehicles.Sincecarsarerestrictedwithrespecttochangeofdirectionandlateralmovementisnotpossible,an’effectivefactor’isincludedforsocialinteractionforcesasaformfactorterm.The‘effectivefactor’variestheinfluenceofforcesexertedfromdifferentdirectionsanddistinguishesbetweenacar-pedestrianoracar-carinteraction.Further,thereisarelationshipbetweenthesteeringangleandthevelocityofavehicle.Therefore,turningconstraintsaregiventoavoidsharpturningbehavioursfordrivers.Therearemoresubtlebehaviouraleffectsaswellthatinfluencedriverresponses.Forinstance,vehiclestendtopasseachotherontheleft(intheUK)ortendtoqueueratherthanovertakeincongestion.Theseaspectsareincluded

byassigningbehaviouralrulesthatadjustparametersoftheforces.Themodelforpedestriansisalsomodifiedastheyarenotonlyinteractingwithotherpedestriansandboundariesbutalsowithvehicles.Further,apredictor-correctormethodcalledGearalgorithmisusedinordertonumericallysolvethesetofordinarydifferentialequationsovertimeforeachmobileobject.Thismethodcandealwithforcesdependentonvelocityandposition.TheGear’sfourthorderpredictoralgorithmkeepstwohigherderivatives to be able to get a better estimation of the new position and velocity.

Havingintroducedthismicroscopicmathematicalfoundationwiththelocalmotionbasedonsocialforces,atacticalalgorithmisadded.Thistacticalleveldetermineswalkingordrivingpathsthathumanbeingsarelikelytochooseundersharedspaceconditions.Infact,theDistancePotentialFieldis generated separately for pedestrians and drivers to indicate the trajectory of the shortest path to reachthedestination.FloodfilldynamicsareusedtocalculatetheDistancePotentialFieldandacombinationofManhattanMetricandChessboardMetricisusedtocalculatedistancesconsideringallobstacles.TheDistancePotentialFieldisre-calculatedforsharedspacewhilethesimulationruns,eachtime a vehicle influences the paths of pedestrians within the shared area.

Inconclusion,basedontheproposedmathematicalfoundationandtheadditionaltacticallevel,ac#simulationforsharedspacesiscreatedwhichiscapableoffacilitatingaquantitativepredictionoftravel patterns. It is shown that the concept of social forces is applicable for modelling the observed movementsofsharedspaceusers.Inaddition,thedistanceandthelastdirectionofmovementofsharedspaceusersprovideenoughinformationtoexplaintheirdecisionmakingprocess.Themodeloffersabasisfortheunderstandingofsharedspaceuserbehavioursanditsrelationshiptospacedesignandtrafficmanagement.

Frozen shuffle update in simple geometries: A first step to simulate pedestrians

Julien Cividini, University Paris-Sud, Paris FRANCECecile Appert-Rolland, University Paris-Sud, Paris FRANCEHenk Hilhorst, University Paris-Sud, Paris FRANCE

Motivatedbyinterestinpedestriantraffic,weintroduceanewtypeofupdateforcellularautomatamodels,thatcouldbeappropriatetomodelpedestriansinparticularatlowormoderatedensities.Inafirststage,wehaveappliedittoasimplemodelusingTASEPlanes.

ATASEP(forTotallyAsymmetricSimpleExclusionProcess)isacellularautomatoninwhichparticlesmoveonalatticebyjumpingfromoneboxtothenextone,alwaysinthesamedirection.Thisisanexclusionprocess,sotherecanbeonly0or1particleoneachsite.HeretheTASEPdoesnothavetobetotallydeterministic,themovementsofthepedestrianscanoccuratsomeprobabilitydifferentfromone,forinstance.Thepositionsoftheparticlescanbeupdatedusingaparallelscheme(theyallmoveatthesametime)orasequentialone(theymoveoneafteranother).Inthatcase,theupdatingordercanbechosentofollowvariousrules.Wedecidedtointroduceanewtypeofupdatescheme.Itiscalled‚frozenshuffle‘update,anditconsistsinupdatingtheparticlesinarandomlychosenorder,whichdoesnotchangeateachtimestepbutisfixedonceforall.Thedynamicsisthendeterministic once the order has been chosen.

Thisavoidspriorityissuesbetweenpedestrians:thefirstpedestriantomovewillsimplyjumponitstargetbeforetheotherone.Thisupdatealsolowersthestatisticalfluctuations,sinceallthepedestrianshavebarelythesamevelocitywhentheyarenotconstrained.Besides,inthefreeflowphase,thereexistsadirectmappingbetweenthediscreteTASEPwithfrozenshuffleupdateandamodelofmovingrodsincontinuousspaceandtime.Weinvestigatethismodelanalytically,andbyMonte-Carlosimulation,fordifferentboundaryconditionsandgeometries.

First,westudyitontheone-dimensionalring,thesimplestnon-trivialgeometry.Indeed,afteraveragingoverallpossibleorders,thereareonlytwomacroscopicparametersleft:theparticledensityandtheparticulecurrent.Wefindthatthereisaphasetransitionforsomecriticalvalueofthedensity:theflowisfreeforsmalldensities,whereasjammingappearsforhigherones.Weareabletoplotthefundamentaldiagram(particlecurrentasafunctionofdensity),andtopredictitanalyticallyforaninfinitesystemandcomputefinite-sizecorrectionsnearthetransition.

Secondly,weuseopenboundaryconditionsfortheTASEPlane,fixingtheentranceandexitprobabilities.Sincewenowhavetocreateparticles,wemustnowinsertthemintothealready-existingparticlesupdatechain.Weusetheaforementionedequivalenceofthemodelwithacontinuousmodelofpedestriansevolvinginacontinuousspace-timetoprescribehowparticlesareinjectedinthesystem.Weobtainnumericallythebulkdensityversustheentranceprobabilityandagainthereisaphasetransition between free flow and jammed flow.

Itturnsoutthatthistransition-whichcanbepredictedanalytically-occurswhentheentranceandtheexitprobabilitiesareequal,asusualforotherupdateschemes,thoughitisnotsostraightforwardhere.Indeed,incontrastwithotherupdateschemes,thecurrentinthejammedphasedependsnotonlyontheexitrate,butalsoontheentrancerate(asthelatterhasaneffectontheupdateorder).

Wehaveconsideredonlythedeterministicversionofthemodel,forwhichpedestriansmovewithprobability1whenitispossible.Theusualmaximal-currentphaseisthereforereducedtoapoint.Weuseadomain-wallapproachtocomputethedensityprofile,whichgivesusthecorrelationlengthand some scaling law near the transition.

Finally,weconsiderasetoftwoopenTASEPlanessharingonebox,inotherwordsacrossing.Wecanthenvarythetwoentranceprobabilitiesandthetwoexitprobabilities,andeachlanecanbeeitherinafreeflowregimeorjammed.Thesimplestcaseiswhenthesystemissymmetricalbetweenthetwolanes,butwealsostudiedanasymmetricsystemwithexitprobabilitiessettoone(inordertohaveonlythejamscreatedbythecrossingandnotbyexitconditions).Eventuallywevaryeachparameterindependently.Withthisupdatescheme,weseethattheparticlestendtoformgroupsthatwecallplatoon,inwhichparticlesallmoveasawhole.Wealsoshowthatapairingmechanismbetweentheplatoonstakesplaceatthecrossingbox,enablingustocomputethephaseboudariesforlargesystems,whicharecoherentwithnumerics.Doingthis,wealsoderivesomeotherfeatures,suchasthedensityand the exit current in each lane.

Infurtherwork,weplantotakeintoaccountthewidthofthecorridorsthroughmultilanemodels,andtoenlargethecrossingtoabiggersquare.Ouraimwouldbetounderstandhowmacroscopicstructurescanspontaneouslyemergeinsuchsystems,andhowtheyaremodifiedforvariousmodificationsofthedynamical rules.

REFERENCES

C. Appert-Rolland, J. Cividini and H. Hilhorst. Frozen shuffle update for an asymmetric exclusion process on a ring. J. Stat. Mech., P07009, 2011.

C. Appert-Rolland, J. Cividini and H. Hilhorst. Frozen shuffle update for deterministic totally asymmetric simple exclusion process with open boundaries. J. Stat. Mech., P10013, 2011.

C. Appert-Rolland, J. Cividini and H. Hilhorst. Intersection of two TASEP traffic lanes with frozen shuffle update. J. Stat. Mech., P10014, 2011.

Comparison of evacuation simulation models

Burkhard Forell, Gesellschaft für Anlagen und Reaktorsicherheit, Köln GERMANYHubert Klüpfel, TraffGo HT GmbH, Flensburg GERMANYSören Schelter, Universität zu Köln, Köln GERMANYVolker Schneider, IST GmbH, Frankfurt am Main GERMANY

Inthispaper,differentfiresafetyengineeringmethodsforthecalculationofegresstimesareapplied.Sincethemodelsarewidespreadanddocumentedindetailelsewhere,theycanbeassumedtoberepresentativeforthefieldofegressmodelling.Therefore,thefocushereisonthecomparisonofthesimulationresults.Theexampleofapplicationisanauditorium.

Onepillaroffiresafetyengineeringarecomputerbasedmethodsforegresssimulation.Whenassessingvariationsforcomplexorexistingbuildings,prescriptiverulesareoftennotsufficienttocoverallrelevantaspectsanddetails.Thisisoftenthecaseforbuildingslikeairports,shoppingcenters,stadia,etc.Thebasicideaistoproofthattheavailablesafeegresstimeissufficient,i.e.therequiredtimeissmaller.Inaddition,theremustbenothreattopersonalsafetyduringtheevacuationprocess.Theavailablesafeegresstime(ASET)isoftendeterminedbasedonCFDsimulationsorderivedfromgeneralconsiderations.Therequiredsafeegresstime(RSET)consistsofatleastfourphases:

RSET=Detection+Alarm+Reaction+Movement

Sincethefocusisonthecomparisonofthemodel,thefirstthreephasesarenottakenintoaccountandthefocusisonthemovementphase.Ifonewishestotakeintoaccountdetecationandalarm,thesetimescanbeaddedtothemovementtime.Thereactiontimecanbeexplicitlyputintothesimulation models (i.e. as a distribution of individual reaction times) but is also not considered for our comparison.

In this contribution we show a comparison of evacuation simulation results for the following models: Aseri,buildingExodus,FDS+Evac,PedGo.TheseresultsarecomparedtothecalculationsbasedonacapacityanalysisandthemodelbyPredtetschenskiiandMilinski.Insummary,sixdifferentmethodsare compared to each other.

Inadditiontothetheresultsfortheevacuationanalysisofthebuilding(auditorium),twosimpleescaperouteelementsweresimulated:(1)a50mhallwayofwidth2mand(2)ahallway10mlongand2mwidewithanascendingstair.Inthefinalsection,criteriafortheassessmentofegresstimesandcongestion are presented.

Theauditoriumtobeevacuatedis34 mx29 mx12 m(LxWxH)andhas20rowswith32seatseach.Upto640personsseatingand360standingmightoccupythevenue.Therearetwostairsinthemiddleandtwostairsattheside.Theauditoriumhastwomainentrancedoorswhichlead(incaseofegress)tothefirstfloorofthebuilding.Theegressrouteisfurtherviaonestairtothegroundfloorandoutside.

Theevacuationtimeisthetimewhenthelastpersonhaslefttheroomviaeithertheemergencyexitsor via the stair leading to the ground floor. I.e. for the escape route via the main entrance doors and the stairtothegroundfloor,thetimeistakenwhentheendofthestair,i.e.thegroundfloorisreached.

TheescaperoutehavebeendesignedaccordingtotheGermanguidelines(Versammlungsstättenverordnung).Accordingtotheseguidelines,thetotalwidthfor1000personsisrequiredtobe6m.Thetwomainentranceandexitdoors(onthefirstfloorleadingtoastaircaseinthefoyer)havetwometerseach.Therearetwoadditionalemergencyexitsatthefrontoftheauditorium(onthegroundfloor)with2meach.Therequirementconcerningthemaximumlengthtoasafeareaisalsofulfilled.Inthesimulation,theemergencyegressviaescaperoute1(mainentrance)andescaperoute2(emergencyexits)iscompared.Reactiontimesarenotset.Theresultsforthefoursimulation programs are compared to the capacity analysis and the calculations based on the model of PredtetschenskiiandMilinski.

Theegresstimesforescaperoute1areinarangeof± 13 %aroundthemean,forescaperoute2from– 16 %to+ 12 %aroundthemean.

Model escaperoute1 escaperoute2 Time Congestion Time CongestionCapacityAnalysis 304s maindoor 298s emergencyexitPredtetschenski 295s internalstair 318s internalstairbuildingEXODUS 382s internalstair 266s internalstairPedGo 348s int.st.,exit 276s int.st.,mainexitFDS+Evac 373s int.st.,exit 239s internalstairASERI 311s internalstair 311s emergencyexitMean 388s +/-13% 285s +16%-12%

Theevacuationanalysishastwomajoraims:(1)determinetheoveralltime(RSET)and(2)identifycritical conditions during an evacuation.

Thetimesobtainedinthedifferentmodelsaredeviatingabout+-20%fromthemeanvalueforallsixmethods and are therefore acceptable.

Theidentificationofcongestionismorediverse,i.e.differentmodelsidentifydifferent„hotspots“.Fortheanalysisofsingleescaperouteelementsthedeviationsoftheresultsfordifferentmodelsaremore prominent than for the complete building.

Humanbehaviorinstressfulsituationsisevenhardertoforetell.Therefore,totheknowledgeoftheauthors,therearecurrentlynodetailedsystematicandquantitativeassessmentcriteriaforasafeevacuation.Severalcriteria,likecongestionanddelaytimes,maximumdensitiesandpressures(orforces)aresummarizedinthispaper.

Poster exhibition

Unsorted list

Maria Davidich, Siemens AG, München GERMANY Waiting zone for real life scenarios. Example a German railway station

Christian Eilhardt, Universität zu Köln, Köln GERMANY Phase separation in pedestrian traffic: An approach to modelling

Sebastian Kühn, Technische Universität Kaiserslautern, Kaiserslautern GERMANY Evacuation dynamics influenced by spreading hazardous material

Jan Peter Ohst, Technische Universität Kaiserslautern, Kaiserslautern GERMANY Minimizing the costs of evacuation paths by decomposing network flows

Jan Dijkstra, Eindhoven University of Technology, Eindhoven NETHERLANDS Modelling time duration of planned and unplanned store visits in a multi-agent simulation of pedestrian activity in city centres

Jun Zhang, Bergische Universität Wuppertal, Wuppertal GERMANY Empirical relations for bidirectional pedestrian stream in a corridor

Stefan Holl, Forschungszentrum Jülich GmbH, Jülich GERMANY HERMES – An evacuation assistant for large arenas

Sebastian Burghardt, Bergische Universität Wuppertal, Wuppertal GERMANY Analysis of flow-influencing factors in mouths of grandstands

Waiting zone for real life scenarios. Example a German railway station

Maria Davidich, Siemens AG, München GERMANYHermann Mayer, Siemens AG, München GERMANYAlexander Pfaffinger, Siemens AG, München GERMANYChristian Royer, Siemens AG, München GERMANY

Pedestrianstreamsimulationsimitatereallifecrowdbehavior,aimingtoreproduceitascloseaspossible.Theirapplicationsrangefromplanningspecialevents,suchasconcertsandsports,collectingexperiencewithcriticalsituations,suchasevacuation,toevaluatingtheplacementofadvertisements.Theutilityvalueofasimulatordependsonhowwellitreproducesrealbehaviourindifferentsituationsthat can occur.

Inmostsimulations,pedestriansare“singleminded”andonlymovetowardstheirtargets.However,ourobservationsonamajorGermanrailwaystationduringseveralfieldexperimentsshowthatasignificantpartofpedestriansdonotconstantlymovetowardsthetargets,butstandstillwaitingforotherpeopleortheirtrainstoarrive.Thesewaitingpersonstherebyinfluencethecrowddynamics,sometimesevenmorethantheoneswhoarewalking.Forexample,atypicalobservedsituationisthefollowing.OnFridayeveningmanypersonscometoarailwaystationandwaitfortheirfriendsandrelativeswhoarecomingbackhomeorcomingforavisit.Usually,theycrowdattheendsofrailwaytracks.Thisisacriticallocationatashorttimeperiodafteratrainarrivalwhenmanypassengersdisembarkatrain.Bystayinginthesecriticallocationsthewaitingpersonsblocktheoneswhoarrivedwithatrain,whichresultsinbottlenecks.Thereby,thestandingpedestrianssignificantlyinfluencethedynamicsofacrowd.Inordertobeabletoreproducethesekindofsituations,pedestrianstreammodelsshouldconsidernotonlymovingpedestrians,butalsopedestrianswhicharestandingor,atleast,notmovingforcertainperiodsoftime.

Here,weproposeamodelforrepresentingwaitingzonesappliedtopedestrianstreamsimulationbasedoncellularautomata.Themainideaofacellularautomata-basedmodelis:Theobservedareaissplitintocellsandeachcellcaneitherbeoccupiedbysomeperson,obstacle,targetorasource,orbeempty.Weimaginethatforcesofattractionexistbetweentargetsandpedestrians,andthatpedestriansarerepulsedbyobstaclesandotherpedestrians.Theseforcesbetweenpedestrians,targetsandobstaclesareexpressedthroughsuitablescalarfunctions:thepotentials.Withthisapproach,oneachtimesteppedestrians chose their direction of movement so that their potential decreases.

Weimprovethismodelbyintroducingwaitingareas.Thewaitingareamodelenablestomodelpedestrianswhonotonlywalktowardstheirtarget,butalsokeepstanding.Assoonasapedestrianhasreachedsomepre-definedwaitingzone,he/shemaychooseanarbitrarylocationwithinthisareaandstand there for a determined period of time.

WevalidatetheproposedmodelonasetofcomplexreallifescenariostakingplaceonamajorGermanrailwaystationduringdaytime.Theobservedbehaviouristhattherearealwayssomepedestriansthat

arewalkingtowardsrailwaytracksandsomethatarewaitingforatrain.Therearealsopedestriansthatare waiting in other parts of the station.

Anobservationsurveycombinesrecordedvideo,extractedfootageandanumberoffieldobservationsonthisrailwaystation.Thevideoshavebeenrecordedduringpeakhoursinthemorningandintheevenin.Severalcamerasrecordedpedestriansinsomepartofthisrailwaystationfromabird‘seyeview.Tracesofindividualpedestriansintimeandspacehavebeenextractedmanuallyfromthevideosusingatoolthatallowstomanually„click“theirpositionsonascreen.Usually,apositionofapedestrianwastakenasthatofhis/herhead.Atthispointwehavefinishedanalyzingtwovideorecordings,eachhavingadurationofatleast1.5minutes.Thenumberofpedestriansobservedoneachvideoisabout400.Theareacoveredoneachvideoincludesseveralplatformsandapartofthestation‘smainhall.Alltrajectorieswithinthecompleteareaofvideoobservationwereextractedandanalyzed.Thetrackingofpedestriansprovidedinformationonvelocitiesdistributionanddensity-velocityrelation.

Ourfieldexperimentsincludedtwoandahalfdaysofobservingthetrainstationfromearlymorningtolateevening.Duringtheseexperiments,thestatisticsonpedestrianarrival/departurerates,pedestrianwalkingtimesateachmeasurementsectionandpedestrianwaitingtimesateachwaitingareawerecollected.Thecollecteddatawasthenusedtocalibratethesimulationmodelsothatthesimulated dynamics reproduces the observed one.

Wedemonstrateseveralreallifescenariosandtheinfluenceofwaitingareasonageneraldynamicsofacrowd.Wealsoshowhowthedevelopedmodelcanbeusedasaplanningtoolandsimulateseveralscenarioswherethesituationbecomescriticalduetowaitingpersonsblockingthewayofothersandtherebycreatingbottlenecks.Thus,thewaitingzonesplayasignificantroleforpedestrianstreamsimulation,andvariouscriticalsituationscannotbeadequatelysimulatedwithouttakingthewaitingzonesintoaccount.

Phase separation in pedestrian traffic: An approach to modelling

Christian Eilhardt, Universität zu Köln, Köln GERMANYAndreas Schadschneider, Universität zu Köln, Köln GERMANY

Theseparationofpedestriantrafficintotwophases(notmovingandslowlymovingpedestrians,respectively)isaneffectthatcanbegenerallyobserved,buthasbarelybeenstudiedsystematically.Theunderstanding of this phase separation deepens the general understanding of pedestrian dynamics and couldthereforeimprovethedesignofairports,shoppingcentersandothercrowdedareasaswellasimprove safety during evacuations.

Theoccurrenceofphaseseparationisacommonandwell-understoodfeatureinvehiculartraffic[1].Therearetwodistinctphases:Ajammedphasewithhighdensityandafree-flowphasewithcarsmovingattheirdesiredvelocity.ThiscanbereproducedinappropriatemodelssuchastheVDRmodel[2].Byusinga”slow-to-start”ruletheoutflowofthejamisreducedcomparedtothemaximalflowofthesystem.Thisleadstoaregionofnon-interactingcarsandthustothefree-flowphase.Theexistenceof two distinct phases is connected to the existence of metastable states in the fundamental diagram. Forintermediatedensityvaluestheflowisnotauniquefunctionofthedensity:thefreeflowbranchcanspontaneouslybreakdownintoacongestedstate.Thisiscalledacapacitydropandleadstoahysteresisloop[3].

Thesituationinpedestriandynamicsismorecomplicatedforseveralreasons.Themostobviousoneisthegenerallytwo-dimensionalnatureofpedestrianmovementincontrasttotheone-dimensionalmovementinvehiculartraffic.Furthermore,collisionavoidanceislessofanissueforpedestrians–pedestriansbumpintoeachotherregularly,carsdonot.Finally,themovementofpedestriansisatleastpotentiallyinfluencedbymultipleotherpedestrianswhereastheinteractionrangeinvehiculartrafficisvery small.

Thereforeexperiments[4]showasimilar,butsomewhatdifferentbehaviorforpedestrianscomparedtowhatisobservedinvehiculartraffic.Firstly,thepedestrianfundamentaldiagramdoesnotshowmetastablestatessimilartothevehicularfundamentaldiagram.However,thetrajectoriesoftheone-dimensional“single-file”pedestrianmovementfeaturetwoseparatephases,namelyajammedhigh-densityphaseandaphaseofmediumtohighdensitywithslowlymovingpedestrians.Toanalyzepedestrianphaseseparation,onethereforehastostudymicroscopicquantities(trajectories),itdoesnotsufficetostudythe(macroscopic)fundamentaldiagram.Incontrasttovehiculartraffic,thedistancebetweenpedestriansinthemovingphaseissmall,notallowingthemtomovewiththeirdesiredvelocity.Bothphasesconsistofinteractingparticles(pedestrians).Themechanismcreatingthephaseseparationthereforediffersfroma”slow-to-start”rule.Therelativelysmalltime-scaleaswellasspatialscale of the actual measurement do not allow to judge the stability of the phase separation.

Wetrytounderstandtheemergenceofthiskindofpedestrianphaseseparation.TypicalpedestrianmodelssuchasthesocialforcemodelbyHelbingetal.[5]orthefloorfieldcellularautomatonmodelbyBursteddeetal.[6]cansuccessfullydescribequalitativephenomenalikelaneformationandareevenusedforquantitativepredictions.However,theydonotfeaturephaseseparation!Forthatreason,we develop a simple cellular automaton model that aims at reproducing the observed phase separation. Thetransitionprobabilitiesofthemodeledpedestriansingeneraldependontheircurrentvelocitiesandontheoccupancyofthenexttwocellsinfrontofthem.Themodelusesaparallelupdate.

Theresultingtrajectoriesofthepedestriansclearlyfeaturetwodistinctphases,namelyacompletelyjammedphaseandaphaseinwhichpedestriansareslowlymoving.Thisphaseseparationisveryslowlydecayingintoasinglecongestedhighdensitystatewithslowlymovingpedestrians.However,the results are in good agreement with the experimental data when observed over a corresponding timescale.Westudyseveralvariantsofthemodelandperformsimulationrunswithvaryingpedestriandensities,maximalvelocitiesandstartingconditions(jammed,homogeneous,random).Furthermore,we discuss possible mechanisms such as anticipation that may generate the experimentally observed phaseseparation.Theanalysisoftheresultingtrajectorieswillbepresentedhere.Finally,wediscussthegeneral feasibility of cellular automaton models to simulate pedestrian phase separation.

REFERENCES

[1] A. Schadschneider, D. Chowdhury, and K. Nishinari, Stochastic transport in complex systems: From molecules to vehicles (Elsevier, 2010)

[2] R. Barlovic, L. Santen, A. Schadschneider, M. Schreckenberg, Metastable states in cellular automata for traffic flow. European Physical Journal B 5, 793 (1998)

[3] F.L. Hall, B.L. Allen, and M.A. Gunter, Empirical analysis of freeway flow-density relationships. Transp. Res. A, 20:197 (1986)

[4] A. Seyfried, A. Portz, A. Schadschneider, Phase Coexistence in Congested States of Pedestrian Dynamics. Lecture Notes in Computer Science 6350, 496 (2010)

[5] D. Helbing and P. Molnár, Social force model for pedestrian dynamics. Phys. Rev. E, 51:4282, (1995)

[6] C. Burstedde, K. Klauck, A. Schadschneider, and J. Zittartz, Simulation of pedestrian dynamics using a 2-dimensional cellular automaton. Physica A, 295:507 (2001)

Evacuation dynamics influenced by spreading hazardous material

Sebastian Kühn, Technische Universität Kaiserslautern, Kaiserslautern GERMANYSimone Göttlich, Universität Mannheim, Mannheim GERMANYJan Peter Ohst, Technische Universität Kaiserslautern, Kaiserslautern GERMANYStefan Ruzika,Technische Universität Kaiserslautern, Kaiserslautern GERMANYMarkus Thiemann, Technische Universität Kaiserslautern, Kaiserslautern GERMANY

Inthisarticle,anevacuationmodeldescribingtheegressincaseofdangerisconsidered.Thereforeweshowhowtocouplethespreadingofsomegaseoushazardousmaterialtotheunderlyingevacuationmodel,whichisbasedoncontinuousnetworkflows.

Weusecontinuousnetworkflowsaswewanttoconsiderregionalevacuation.Thereforeweliketomodeltheevacueesasahomogeneousgroupanddonottakeindividualbehaviorintoaccount.Thuswearegoingtomakeuseofamacroscopicmodel.Weusepartialdifferentialequationstomodelthetransportonthearcsofthenetworkandmakeuseofsomelinearityassumptionsontheflowfunctiontobeabletodiscretizethecontinuousmodelinastraightforwardway.Additionally,forthecorrectformulationoftheproblemusingpartialdifferentialequationsitisnecessarytodefinecouplingconditionsatnodesensuringtheconservationofmass.Forinstance,thisisdoneintrafficflow[2,6]orsupplychainmodelling[4,3].Especiallytheformerareclosetoevacuationmodellingandthusmanyideascanbeappliedhere,too.

Weareabletoshowthatthecontinuousnetworkflowmodelweareusingcanbediscretizedasadiscretedynamicnetworkflowproblem.Herewehavetomentionthattheabilitytousediscretedynamicnetworkflowsasdiscretizationforourcontinuousmodelstronglyreliesonthelinearityassumptionsontheflowfunction.Butbecauseofthisspecialdiscretizationweareabletostateanefficientalgorithmalwaysleadingtotheglobaloptimum,whereoptimalitycantherebybeunderstoodwithrespecttotwodifferentmeasures:fastestegressandsafestevacuation.

Ontheonehandfastestegressisasuitableobjectivefunction,sincethezoneofdangershouldbeclearedintheshortestpossibletimetoavoidtheexposureoftheevacueestothehazardousmaterialbyleadingthemawayfromitssourceasfastaspossible.Thisobjective,alsoknownasthequickestflowproblem,hasbeenwidelystudiedfordiscretedynamicnetworkflowsforexamplebyBurkhardetal.[1]andHamacheretal.[5],wherethelaterhasalsodealtwithtime-dependentcapacities,enablingmodelswithtemporallyfailureofarcs.Flowandloaddependenttraveltimes,reflectingtheslow-downofevacueesduetocongestionofarcsareconsideredbyKöhleretal.[7,8].Sincewediscretizeourcontinuousnetworkmodelinsuchawaythatwecanreformulateitasadiscretedynamicnetworkflowmodelwecanmakeuseofthetechniquesmentionedaboveandarethereforeabletostateanefficientalgorithm.

Ontheotherhandweconsidersafestevacuationasanobjective.Thismeansthatwetakethespreadingofthehazardousmaterialintoaccountwhenplanningtheevacuation.Asthisspreadingisnotnecessarilyuniformlydistributedaroundthesourceofthehazardthisobjectiveyieldsdifferent

solutions than the one discussed before.

Thespreadingofthehazardousmaterialisdescribedusingalinearadvection-diffusionequationsinceweliketofocusonthemappingofthehazardonthenetworkinthisfirstapproachasamorerealisticpropagationmodelmaybeincludedinalaterstep.Themappingofthehazardousmaterialontothenetworkisthendonebyintroducingaspecialcostfunctionrepresentingthelevelofhazardtowhichtheevacueesareexposed.Thesecostsare-incontrasttothetraveltimesconsideredinthecaseoffastestegress-time-dependent.Tosolvethecoupledmodelforsafestevacuationweuseagainthediscretizationasadynamicnetworkflowproblemensuringthefindingofaglobaloptimumbyusingslightly other algorithms than for fastest egress.

REFERENCES

[1] R.E. Burkard, K. Dlaska, and B. Klinz, The quickest flow problem, Methods and Models of Operations Research, 37 (1993), 31-58.

[2] G.M. Coclite, M. Garavello, and B. Piccoli, Traffic flow on a road network, SIAM Journal on Mathematical Analysis, 36 (2005), 1862-1886.

[3] C. D‘Apice, S. Göttlich, M. Herty, and B. Piccoli, Modeling, Simulation and Optimization of Supply Chains: A Continuous Approach“, SIAM book series on Mathematical Modeling and Computation, 2010.

[4] S. Göttlich, M. Herty, and A. Klar, Network models for supply chains, Communications in Mathematical Sciences, 3 (2005), 545-559.

[5] H.W. Hamacher and S.A. Tjandra, Eariest arrival flows with time dependent capacity for solving evacuation problems, in „Pedestrian and Evacuation Dynamics“ (eds. M. Schreckenberger and S.D. Sharma), Springer, Berlin, (2002), 267-276.

[6] M. Herty and A. Klar, Modeling, simulation, and optimization of traffic flow networks, SIAM Journal on Scientific Computing, 25 (2003), 1066-1087.

[7] E. Köhler, K. Langkau, and M. Skutella, Time-expanded graphs for flow-dependent transit times, Lecture Notes in Computer Science, 2461 (2002), 599-611.

[8] E. Köhler and M. Skutella, Flows over time with load-dependent transit times, SIAM Journal on Optimization, 15 (2005), 1185-1202.

Minimizing the costs of evacuation paths by decomposing network flows

Jan Peter Ohst, Technische Universität Kaiserslautern, Kaiserslautern GERMANYStefan Ruzika, Technische Universität Kaiserslautern, Kaiserslautern GERMANY

Dynamicnetworkflowmodelsareausefultooltocalculatemacroscopicevacuationplansforlargescaleregionalevacuation.Insuchmodelseveryroadisrepresentedbyanundirectededgeinanetworkwherethenodesarethejunctionsbetweenroads.Toeveryedgeamaximalinflowcapacityisassignedas well as a time needed to travel on the edge from start to end.

Overtheyearsmanymodelsandalgorithmshavebeendevelopedtocontributetovariousfacetsoflargescaleroutingandevacuationproblemsovertime.Thisincludestheoptimizationoftotalevacuationtime,earliestarrivalofevacueesandmaximumthroughput[1].Themodelscanalsobeextended by introducing additional cost functions on the edges for example to model a scenario where an evacuation happens in the presence of a toxic cloud. In such scenarios not only the evacuation timeliesinthefocusoftheoptimizationbutalsothecostsofthechosenevacuationplan.Intheaboveexamplethiscorrespondstominimizingtheexposuretothetoxinduringtheevacuationprocess.

Thesolutionofthenetworkflowproblemsisnaturallyaflowonanetwork.Howevertodecidewheretheevacueesshouldbesentadecompositionofthisflowintoevacuationpathsisrequired.Whileitisstraightforwardtofindanarbitrarydecompositionofaflowthisdecompositionmightnotbeunique.Inadditionnoteverydecompositionwillbeequallygoodfromapracticalpointofview.Infactthechoiceofaspecificdecompositionmighthaveatremendouseffectonthequalityofthesolution.Forexampleintheabovescenariowheretheevacuationtakesplaceinthepresenceofatoxiccloudaflowminimizingthetotalexposureoftheevacueestothehazardmighthavedifferentdecompositions.Inoneofthemeverypersonmightbeexposedtonearlythesameamount,whileinanotherdecomposition some evacuees are exposed to an unacceptable high degree while other are not exposed atall.Notethatthosetwosolutionsareequivalentintermsofthetotalexposuretothetoxinbutareobviouslyverydifferentinquality.

Thequestionconsideredhereishowanoptimaldecompositionofagivenflowwithrespecttosomecostfunctioncanbefound.Thisisaproblemthathasbeenaddressedveryrarelyinliteraturebutstillisverycrucialforevacuationplanning.IthasbeenshownthattheproblemisNP-hard[2]andhenceitcannotbeexpectedtofindalgorithmsthatsolvetheprobleminpolynomialtime.Nevertheless an integer program is presented which can be used to compute optimal decompositions of small instances. For larger instances however both the number of constraints in the program as well as the runtimeincreasetoapointwhereitisnotlongerefficienttousethisapproach.Forthisreasonalsoaneasytoimplementapproximationalgorithmispresentedwhichisbasedonagreedystrategy.Thisapproximationalgorithmcanalsobeenusedforlargescalenetworksbutmightnotreturnthebestsolutionpossible.Thequalityofthosesolutionsisanalyzedandcomparedtotheoptimalsolutions.

Obviouslythebestsolutionofthedecompositionproblemisafaironewhereallpathshavethesamecosts.Howeversuchasolutiondoesnotnecessarilyexistfortheflowwehavechosen.Infactitispossibletoconstructscenarioswhereitisonlypossibletofinddecompositionsthathavehighcostsonsomepathsandlowcostsonothers.Sostartingfromanarbitrarynetworkthequalityofadecomposition not only depends on the way a given flow is decomposed but also on the flow itself. Itseemsreasonablefortheproblemgiventochooseaflowthatminimizesthetotalcostbutthereisnoguarantythatthisistheflowthathasthebestdecomposition.Solvingthemoregeneralproblemstartingfromanarbitrarynetworkisevenharderthantheproblemconsideredhere,sincefindingadecompositionofagivenflowarisesasanaturalsubproblemofthetotalproblem.Howeveritcanbestudied how the structure of the flow that is decomposed as well as the costs of the edges used by this flowaffectthequalityoftheapproximationalgorithmaswellasthelowerandupperboundsoftheoptimalsolution.Thisknowledgethenmightbeusedtoconstructagoodstartingflowforwhichthedecomposition is calculated.

REFERENCES

[1] H. W. Hamacher and S. A. Tjandra. Mathematica modelling of evacuation problems: A state of the art. Berichte des Fraunhofer ITWM, 24:1–38, 2001

[2] Y. Hendel and W. Kubiak. Decomposition of flow into paths to minimize their length. Available online at http://sites.google.com/site/yannhendel/Decomposition.pdf?attredirects=0.

Modelling time duration of planned and unplanned store visits in a multi-agent simulation of pedestrian activity in city centres

Jan Dijkstra, Eindhoven University of Technology, Eindhoven NETHERLANDSJoran Jessurun, Eindhoven University of Technology, Eindhoven NETHERLANDSBauke de Vries, Eindhoven University of Technology, Eindhoven NETHERLANDSHarry Timmermans, Eindhoven University of Technology, Eindhoven NETHERLANDS

Agent-basedmodellingisacomputationalmethodologythatallowsustocreate,analyze,andexperimentwithartificialworldspopulatedbyagents.Aspecificresearchareaismicro-scaleagent-based modelling that can be used for the simulation of pedestrian movement for low and high density scenariosandfortheeffectofchangesintheenvironment.Suchmodelscanalsobeusedforpedestriandynamicsincitycentrestoshowthedesigneffectsintheshoppingenvironment.Therefore,amulti-agentmodeltosimulatepedestriandynamicdestination,routeandschedulingbehaviourisunderdevelopment,wherethesimulationofmovementpatternsisembeddedinamorecomprehensivemodel of activity travel behaviour.

Representationisamainissueinsimulatingpedestriandynamics.Onecandistinguishtherepresentation of the pedestrian environment and the representation of pedestrians. In the domain ofacitycentre,representationofapedestrianenvironmentincludesthegeometryoftheshoppingenvironmentsuchasstoresandstreets,thenetworkasacellulargrid,andpedestrianobjects.Pedestrianrepresentationincludessocioeconomiccharacteristics,speed,goals,familiaritywiththeenvironment,andactivityagenda.Itisassumedthatpedestriansperceivetheirenvironmentandthattheyaresupposedtocarryoutasetofactivities.Forcompletinganactivity,pedestriansspendtimeinstores.Asaconsequence,timedurationinfluencestheirmovementbehaviouroverthenetwork.

Althougha3Dpresentationofpedestriansandthepedestrianenvironmentforthesimulationofpedestrianmovementistheultimategoal,itisneverthelessmeaningfultotesttheunderlyingprinciplesinanappropriate2Drepresentationofpedestriansandtheirenvironment.NetLogocanbeusedasimulationtoolkitbecauseitisasuitablesimulationframeworkthatsupportsmodelling,simulationandexperimentation.Italsooffersskeletonsofagentsandtheirenvironment,andinteroperability(e.g.GIS).Wewilluseshape-fileinformationoftheenvironmentandnetworkstructureforvisualizingthe2DenvironmentandNetLogofortheactualsimulation.Forpopulatingpedestrianagentsintheenvironmentandforattachingactivityagendastopedestrianagents,aMonteCarlosimulationisusedwhich implies that the behaviour of each pedestrian agent is simulated by a series of draws of random numbersfromsuccessiveprobabilitydistributions.Theprobabilitydistributionsarebasedonrealdatacollections,suchastimespentinastore,attachinginnerlaneorouterlaneasanentrypointandpedestriancharacteristics(age,gender,etc.).

Themainsubjectofthispaperistointroducetheimplicationoftimedurationofplannedandunplannedstorevisitswithinasimulationframeworkforpedestrianmovementsimulation.Thisframeworkinvolvesanagent-basedmodelthatprovidesanactivityagendaforpedestrianagentsthatguides their shopping behaviour in terms of destination and time spent in shopping areas. In order to

implementtheactivityagenda,pedestrianagentsneedtosuccessivelyvisitasetofstoresandmoveoverthenetwork.Itisassumedthatpedestrianagents’behaviourisdrivenbyaseriesofdecisionheuristics.Agentsneedtodecidewhichstorestochoose,inwhatorderandwhichroutetotake,subjecttotimeandinstitutionalconstraints.Itisassumedthatpedestrianagentsareindifferentmotivationalstates.Theymayateverypointduringthetriphavegeneralinterestsinconductingparticularactivities,withouthavingdecidedonthespecificstoretovisit,buttheymayalsobeinamoregoal-directedmotivationalstateinwhichcasetheyhavealreadydecidedwhichstoretovisit.Todefinepedestrianagents’motivationalstates,thecategorizationgoal-oriented,leisureshopping,andnospecificintentionwasmadefortheirvisit.Themotivationalstatesareofinfluenceontheimpulseandnon-impulsestorechoiceprocessesandthereforeontheplannedandunplannedvisitstoastore.Alltheseaspectsaffectpedestrianagents’timedurationinvisitingstores.Pedestrianagentsmoveoverastreetnetworkandarepartofapedestrianflowinthisstreetnetwork.Howeverpedestrianagentscanbetemporarilyremoved from the pedestrian flow by visiting a store and participating again in the pedestrian flow aftervisitingthatstore.Inthatcase,thetimespentbyapedestrianagentinastoreisrelevant.ForthesimulationrunthistimedurationisdeterminedbyaMonteCarlosimulation.ThefindingsfromthecollecteddataofthedurationofavisittoastoreindicatethatthistimedurationmeetstheWeibulldistribution,andthatthisdurationalsodependsofthestorecategoryaswellasthepriorityofthestore.

Thispaperpresentsthefindingsoftheimplicationoftimedurationofavisittoastorewithinasimulationframeworkforpedestrianmovementsimulationandthedependenciesofthistimedurationwith planned and unplanned visits to a store.

Empirical relations for bidirectional pedestrian stream in a corridor

Jun Zhang, Bergische Universität Wuppertal, Wuppertal GERMANYWolfram Klingsch, Bergische Universität Wuppertal, Wuppertal GERMANYAndreas Schadschneider, Universität zu Köln, Köln GERMANYArmin Seyfried, Bergische Universität Wuppertal, Wuppertal GERMANY

Inrecentyears,extensivestudiesonuni-andbidirectionalpedestrianflowincorridorhavebeendoneempiricallyandnumerically.However,discrepanciesexistamongpreviousstudiesandthereisnoconsensusabouttheoriginofthesediscrepancies.Evenforthefundamentaldiagram,thebasicrelationbetweendensity,velocityandflowrate,obviousdifferencescanbeobservedespeciallyfordensityρ>2.0m-2whilecomparingfundamentaldiagramsfromdifferentstudies.Moreoverthereisnoconsensuswhetherthefundamentaldiagramsofuni-andbidirectionalflowsaredifferentornot?PredtechenskiiandMilinksii[1]andWeidmann[2]neglectedthedifferencesinaccordancewithFruin,whostatesthatthefundamentaldiagramsofmulti-anduni-directionalflowdifferonlyslightly[3].WhileHelbingetal.[4]concludedthatcounterflowsatbottlenecksaresignificantlymoreefficientthanunidirectionalflows.Besides,theflowratioofpedestrianinoppositestreamsisalsoanimportanttopicinbidirectionalstream.NavinandWheeler[5]foundareductionoftheflowindependenceofdirectionalimbalances.Pushkarevetal.[6]andLametal.[7,8]foundthatbidirectionalflowisnotsubstantiallydifferentfromunidirectionalflowaslongasthedensitiesoftheoppositestreamsarenottoodifferent.However,Olderetal.statethatdifferentratiosofflowinbidirectionalstreamdonotshowanysignificanteffectonthewalkingspeed[9].

Toresolvethesediscrepanciesseriesofexperimentsunderlaboratoryconditionswerecarriedouttostudythecharacteristicsofbidirectionalpedestrianstreamsinacorridor.Twodifferentkindsofbidirectionalstreamsarecreatedinourexperiment:1)pedestrianstreamswithstableseparatedlanesand2)pedestrianstreamswithdynamicmulti-lanes(withbalancedandimbalancedflowratiosseparately).Theexperimentswererecordedbytwocamerasmountedonarackattheceilingofthehall.ThepedestriantrajectorieswereautomaticallyextractedfromvideorecordingsusingthesoftwarePeTrack[10].GiventheadvantageofsmallscatterandhighresolutionintimeandspaceoftheVoronoimethod[11],weuseittocalculatepedestriancharacteristicsincludingdensity,velocityandflowfrom these trajectories. It is shown that the fundamental diagram for various forms of bidirectional streamsagreewellandnolargedifferencesarefoundinobserveddensityrange.Thatistosay,thenumberandformoflanesincorridordoesn’thaveobviousonthepedestrianflowfordensityρ<2.5m-2.Bycomparingthefundamentaldiagrambetweenuni-andbidirectionalstream,itisshownthatthedensity-flowrelationshipsagreewellforthefreeflowstateatdensityρ>1.0m-2.However,thevelocitiesforunidirectionalflowarelargerthanthatofbidirectionalflowforρ>1.0m-2.Themaximumspecificflowinunidirectionalstreamsissignificantlylarger(about0.51/m∙shigher)thanthatinbidirectionalstreams.Inaddition,thejammingtransitionfromfreeflowstatetojammedstate,whichisfoundinmanytheoreticalmodels,wasnotobservedinourexperimentcoveringadensityrangeuptoρ=3.5m-2.

REFERENCES

[1] V. M. Predtechenskii and A. I. Milinskii, Planning for Foot Traffic Flow in Buildings. Amerind Publishing, New Delhi, 1978, translation of: Proekttirovanie Zhdaniis Uchetom Organizatsii Dvizheniya Lyuddskikh Potokov, Stroiizdat Publishers, Moscow, 1969

[2] U. Weidmann, “Transporttechnik der Fussgänger,” Institut für Verkehrsplanung,Transporttechnik, Strassenund Eisenbahnbau, ETH Zürich, ETH Zürich, Tech. Rep. Schriftenreihe des IVT Nr. 90, 1993, zweite, ergänzte Auflage.

[3] J. J. Fruin, Pedestrian Planning and Design. Elevator World, New York, 1971

[4] D. Helbing, L. Buzna, A. Johansson, and T. Werner,“Self-Organized Pedestrian Crowd Dynamics: Experiments, Simulations, and Design Solutions,” Transportation Science, vol. 39, pp. 1–24, 2005

[5] F. D. Navin and R. J. Wheeler, “Pedestrian flow characteristics,” Traffic Engineering, vol. 39, pp. 31–36, 1969

[6] B. Pushkarev and J. M. Zupan, “Capacity of Walkways,” Transportation Research Record, vol. 538, pp. 1–15, 1975

[7] W. H. K. Lam, J. Y. S. Lee, and C. Y. Cheung, “A study of the bi-directional pedestrian flow characteristics at Hong Kong signalized crosswalk facilities,” Transportation, vol. 29, pp. 169–192, 2002

[8] W. H. K. Lam, J. Y. S. Lee, K. S. Chan, and P. K. Goh,“A generalised function for modeling bi-directional flow effects on indoor walkways in Hong Kong,” Transportation Research Part A: Policy and Practice, vol. 37, pp. 789–810, 2003

[9] S. Older, “Movement of Pedestrians on Footways in Shopping Streets,” Traffic Engineering and Control, vol. 10, pp. 160–163, 1968

[10] M. Boltes, A. Seyfried, B. Steffen, and A. Schadschneider,“Automatic Extraction of Pedestrian Trajectories from Video Recordings,” in Pedestrian and Evacuation Dynamics 2008, W. W. F. Klingsch, C. Rogsch, A. Schadschneider, and M. Schreckenberg, Eds. Springer-Verlag Berlin Heidelberg, 2010, pp. 43–54

[11] J. Zhang, W. Klingsch, A. Schadschneider, and A. Seyfried, “Transitions in pedestrian fundamental diagrams of straight corridors and T-junctions,” Journal of Statistical Mechanics: Theory and Experiment, vol. P06004, 2011

HERMES – An evacuation assistant for large arenas

Stefan Holl, Forschungszentrum Jülich GmbH, Jülich GERMANYArmin Seyfried, Forschungszentrum Jülich GmbH, Jülich GERMANYAndreas Schadschneider, Universität zu Köln, Köln GERMANY

Thetrendtowardslargemultifunctionalbuildingsaswellasthedimensionsofpubliceventsmakesnewdemandsonthequalityofsecurityconcepts.Incaseofemergency,alltheattendeesmustbeabletoevacuatethedangerarearapidly.Althoughthisisgenerallyensuredbytheapplicationofbuildingregulations,itcanresultinadangerouscrushandlongqueuesincaseofovercrowdingorifsomeoftheemergencyexitsareblocked.Inrecentyearswehaveseenanincreasingnumberofcrowddisasterswithmanyinjuredpersons,e.g.LoveParadeinDuisburg(Germany).

Inordertopreventsuchcriticalsituationsoptimalcrowdmanagementneedsaccurateandup-to-dateinformationaboutthecurrentstateinthearena.Inthethree-yearresearchprojectHERMES–fundedbytheGermanFederalMinistryofEducationandResearch(BMBF)–wehavedevelopedanevacuationassistant.Intheprojectatotalof13partnershaveworkedtogether.Inadditiontotheresearchinstitutionsandbusinessenterprisestheendusers(police,firefightersandsecurityguards)wereinvolvedintheexplorationoftheevacuationassistant.Theaimoftheassistantistoincorporatethecurrentstateofthebuilding,butalsothenumberanddistributionofpersonspresentinthesimulation.Soforthefirsttimeitispossibletoobtainreal-timesimulationresultsbasedontheactualrisksituation.

TheESPRITArenainDüsseldorf(Germany)providedavenueforimplementingtheevacuationassistant.Theexampleofthismultifunctionalareawithacapacityofmorethan60,000spectatorsshowed how crowds of people at big events can be guided so that optimal use can be made of the emergencyexits.TheaimoftheHERMESProjectistoproduceareal-timesimulationforareliableforecastoftheevacuationbehavior.Thetaskforceswillreceiveimportantinformation,butthedecisionontheoptimalstrategywillbetakenineachcasebytheforcesthemselves.

TheprojectHERMESconsistsofseveralcomponents:a) Safetyandsecuritymanagementsystemb) Video-basedpersoncountingc) Real-timesimulationcored) Communicationmodule

Thesafetyandsecuritymanagementsystemprovidesinformationabouttheconditionofthebuilding.Soweknowifescaperoutesaresmokyorwhetherpartsofthebuildingwereclosede.g.becauseofabombthreat.Thevideo-basedpersoncountingprovidesreliabledataonthenumberofpeoplewhoarecurrentlyinthepartsofthebuilding.Dataprivacyistakenveryseriouslyhere:wedonotstorevideoimages,butonlythecountdatawillbepassed.Basedonthesedata,thesystemoperatorcannowdefine

a scenario that will be calculated from the simulation tools.

InourprojectHERMES,threedifferentmodelsarecombinedtoobtainreliablepredictionsfromthesimulationsandachieveanoptimalperformance.Oneapproachreliesonamacroscopicmodelallowingaroutingrecommendation[1].Additionallyweusetwomicroscopicmodels(floorfieldcellularautomatonmodel[2]andthegeneralizedcentrifugalforcemodel[3])whichprovideratherdetailedpredictionsaboutthedevelopmentofanevacuation.Toincreasethereliabilityoftheforecastlarge-scaleexperimentshavebeenperformedinthelaboratoryandthearenaitself.Thesedataarecomplementedbyfieldstudies.Theresultsfromtheseexperimentshavebeenusedtocalibratethemodelsinordertomakequantitativepredictions.

Essentialforthepracticalityoftheevacuationassistantisthetreatmentoftheresultingdata.Thefrontendofthesystemisthereforeacommunicationmodule.Thetypeofvisualizationhasbeenjointlydevelopedandtestedwiththeendusers.Withatrafficlightsystem(red,yellow,green)itcanbedeterminedonafirstglancewhethertheallowablenumberofoccupantsisexceededinsomeareasofthebuilding.Basedonthemicro-simulationitisshown,wherecriticaltrafficjamswillariseduringthenext15minutes.Duetotheseresultstheresponsibleteamleadershavetheopportunitytodeploytheiremployeessothatpedestriantrafficcanbedirectedoptimally.

In this contribution we will present an evacuation assistant for mass events providing results in real-time.TheassistantiscurrentlyabsolvingatestphaseinthefacilitiesoftheESPRITArenainDüsseldorf.AnoverviewofresultsandconclusionsoftheprojectHERMESwillbepresented.

REFERENCES

[1] A. Schomborg, K. Nöckel and A. Seyfried. Evacuation Assistance for a Sports Arena Using a Macroscopic Network Model. In: R. Peacock, E. Kuligowski and J. Averill (ed.), Pedestrian and Evacuation Dynamics 2010, Springer, p. 389-398 (2011)

[2] C. Burstedde, K. Klauck, A. Schadschneider, and J. Zittartz. Simulation of pedestrian dynamics using a two-dimensional cellular automaton. Physica A, 295:507525, 2001.

[3] M. Chraibi, A. Seyfried, and A. Schadschneider. Generalized centrifugal force model for pedestrian dynamics. Physical Review E, 82:046111, 2010.

Analysis of flow-influencing factors in mouths of grandstands

Sebastian Burghardt, Bergische Universität Wuppertal, Wuppertal GERMANYWolfram Klingsch, Bergische Universität Wuppertal, Wuppertal GERMANYArmin Seyfried, Bergische Universität Wuppertal, Wuppertal GERMANY

Sincemanyyears,footballisoneofthemostpopularsportsinGermany.DuringtheGermanSoccerLeagueseason,everyweekthousandsoffansoccupythegrandstandsoffootballstadiumstowatchandsupporttheirfavoriteteam.Thelargenumberofpeopleensuresagreatatmosphere,butalsoinvolvesrisksconcerningsafetyaspects.Inhazardoussituation,aquickevacuationcanbenecessary.Peoplehavetoleavetheirseatsandfollowtheevacuationroutestothemouthofthegrandstand.Themouthitself,i.e.theconnectiontotheexit,hasakeyfunctionintheevacuationprocess,becausedifferentpedestrianstreamsmerge.Ourgoalistoanalyzemovementofpedestrianstreamsquantitativelyingrandstandsofstadiumswithspecialregardtotheconducttothemouth.Thusthedesignofthegrandstandincludingthelayoutofpathsandthemouthcouldbeoptimizedregardingsafetyandeconomic aspects.

Thegermanbuildingcodeoffersseveralrequirementsconcerningthewidthandlengthofescaperoutes,butthegeometryofthemouthisn’tconsideredyet.Alsoforthemergingofdifferentpedestrianstreams,insufficientinformationisavailableintheliterature.OnlyPredtetschenskiiandMilinskii[1],whopresentamacroscopicmodeltodescribepedestrianmovementquantitatively,givesomeinformation about the merging process. Incoming flows are added and the capacity of the foregoing corridorischecked.Thus,ifthesumoftheincomingflowsexceedsthecapacity,jammingwilloccur.

Toimprovetheknowledgebaseandthelevelofsafetyregardinganevacuationofplacesofassemblylikeasportsstadium,experimentsunderlaboratoryconditionswerecarriedoutintheESPRITarenainDüsseldorfwithintheHERMESproject[2].Overalleightrunswithdifferentsetups(upto300persons)wereperformedinthegrandstandstoanalyzethemergingeffectoftwoandthreepedestrianstreams,theinfluenceofdifferentdensities(e.g.everysecondseatused)andobstacleslikeacloseddoorwingorsecuritypersonnel.Toproofthedependencyoftheinclineofthegrandstandontheseparameters,therunshavebeenperformedinthesub-andupper-rank(incline27°and33°resp.)oftheESPRITarena.Theexperimentswererecordedbytwostereocamerasandprecisepedestriantrajectories(±50mm)wereextractedautomaticallyfromvideorecordingsusingthesoftwarePeTrack[3].Thusamicroscopicanalysisofpedestriancharacteristicsincludingdensity,velocityandflowisfeasible.

Theresultsoftheexperimentsshow,thattheinclineofthegrandstandhasaninfluenceonthepedestrianflowonthestairs,whichleadtothemouth.Thus,theflowinthemouthisaffectedbythereducedincomingflowrespectively.Maximumspecificflowmeasuredinthemouthoftheupper-rankis1.66(m∙s)-1andinthesub-rank1.83(m∙s)-1.ThattheinclineinfluencesthefundamentaldiagramforstairsisalreadydocumentedintheSFPEHandbook[4]andbyFrantzich[5,6],buttheinfluenceontheflowinthemouthconnectedwiththeoverallevacuationtime,isn’tconsideredyet.Our

experimentsshow,thatthestrategicpositioningofthesecuritypersonnelhasasignificantinfluenceonthepedestrianflowinthesub-rank,butsurprisinglyonlyaminorinfluenceontheflowintheupper-rank.UsingtheVoronoimethod[7],wemadetopographicalmeasurementsinthemouthduringstationarystates.Highestdensityappears(ρ≈3m-2)intheregionwheretheincomingpedestrianstreams merge.

REFERENCES

[1] V. M. Predtechenskii and A. I. Milinskii, Planning for Foot Traffic Flow in Buildings. Amerind Publishing, New Dehli, 1978, translation of: Proekttirovanie Zhdanii s Uchetom Organizatsii DvizheniyaLyuddskikh Potokov, Stroiizdat Publishers, Moscow, 1969.

[2] “Hermes - investigation of an evacuation assistant for use in emergencies during large-scale public events,” http://www.fz-juelich.de/jsc/hermes, March 2011.

[3] M. Boltes, A. Seyfried, B. Steffen, and A. Schadschneider,“Automatic Extraction of Pedestrian Trajectories from Video Recordings,” in Pedestrian and Evacuation Dynamics 2008, W. W. F. Klingsch, C. Rogsch, A. Schadschneider, and M. Schreckenberg, Eds. Springer-Verlag Berlin Heidelberg, 2010, pp. 43–54

[4] P. J. DiNenno, SFPE Handbook of Fire Protection Engineering, 3rd ed. Quincy MA: National Fire Protection Association, 2002.

[5] H. Frantzich, “A model for performance-based design of escape routes,” Department of Fire SafetyEngineering, Lund Institute of Technology, Tech. Rep. 1011, 1994.

[6] Frantzich, H., “Study of movement on stairs during evacuation using video analysing techniques,”Department of Fire Safety Engineering, Lund Institute of Technology, Tech. Rep., 1996.

[7] J. Zhang, W. Klingsch, A. Schadschneider, and A. Seyfried, “Transitions in pedestrian fundamental diagrams of straight corridors and T-junctions,” Journal of Statistical Mechanics: Theory and Experiment, vol. P06004, 2011

Poster exhibition

Unsorted list

Beril Sirmacek, Deutsches Zentrum für Luft- / Raumfahrt, Oberpfaffenhofen GERMANY Understanding social behaviors of pedestrian crowds in airborne images based on a graph-cut and sub-graph matching based framework

Sven Hebben, TraffGo HT GmbH, Flensburg GERMANY Evacuation analysis for venues: Systematical approach and comparison to evacuation trials

Konstantinos Kostas, Technological Educational Institute of Athens, Athens GREECE Motions effect for crowd modeling aboard ships

Gabriel Wurzer, AIT Austrian Institute of Technology, Wien AUSTRIA Early-stage egress simulation for process-driven buildings

Xiaolei Zou, Tongji University, Shanghai CHINA Experimental study of the passengers’ self-learning in spatial structure of urban mass transit stations with cognitive map

Douglas Samuelson, InfoLogix Inc., Annandale VA USA Agent-based simulations of pedestrian movement for site security: U. S. Secret Service‘s current capabilities and next steps

Martin Lopusniak, Technical University of Košice, Košice SLOVAKIA Comparison of simulation results with evacuation test and normative calculation

Daniel Weber, Universität Duisburg-Essen, Duisburg GERMANY Merging of pedestrian queues: a mean field approach for CA models

Carlos Pretto, Federal University of Rio Grande do Sul, Porto Alegre BRAZIL Simulation model for vehicle and pedestrian interaction considering road crossing activities

Characteristics of pedestrian trips in Indian cities

Sanjay Gupta, School of Planning and Architecture, New Delhi INDIA

Pedestrianisoneofthemostvitalcomponentsofanytransportationsystem.Thepedestrian,byfar,constitute the most vulnerable road user group comprising high proportion of road accident fatalities .Inordertopromotewalkingasasafe,healthyandsustainableformoftransportandprioritisingthe needs of pedestrians in the urban environment it becomes necessary to manage its movement. In theinnercity,thiswillincludetheidentificationofconflictpointsandthedevelopmentofimprovedmovementinthecontextoftheemergingurbanstructureandurbanframeworkwhileforoutsidethe city core emphasis could be placed which facilitates on creating and promoting a safe pedestrian environment.PedestrianandcyclingpoliciesinadevelopingcountrylikeIndiacouldevenbemoreimportanttodaywhencitiesareplanningmassiveaugmentationofpublictransport.Metro,busrapidtransitsystemandstandardbusescannotworkoptimallyifthesearenotsupportedwithagoodpedestriannetwork.Anyattempttoimprovetheshareofpublictransportwillleadtocorrespondentincreaseinwalkingandroadswillhavetobeplannedwithmorewalkingspace.ThepresentpaperisbasedontheevidencecollatedandanalysedonthepedestriantripscharacteristicsacrossdifferentsizedcitiesinIndia.

ThegrowthofurbanpopulationinIndiahasbeenextremelyrapidduringthecourseoflastcentury.WhilethetotalpopulationofIndiahasgrownfrom361min1961to1027min2001theurbanpopulationhasincreasedfrom62mto285minthesameperiod.Thepercentageshareofurbanpopulationhasgoneupfrom17.3%in1951to27.8%in2001.Thenumberofurbansettlementstoohavegoneupfrom2845in1951to3969in2001.TheshareofClass-Icities(withpopulationofonelakhandabove)inthetotalurbanpopulationhasincreasedfrom44%to68%duringtheaboveperiod.Theconcentrationofpopulationinthe‚millionplus‘citieshasbeenparticularlystriking.Theirnumberhasincreasedfrom5in1951and35in2001.These35citiestogetherhadapopulationof107.1millionin2001,accountingfor37.7%ofIndia‘surbanpopulation.Theaveragetriplengthtendstoincreasefrom3.16kminsizeIcitiesto5.58kminsizeVIcitieswhilethemechanizedtriplengthsincreasefrom4.04kmto7.97kmrespectively.Theshareofpublictransporttripsincreasesfrom18.8%insizeIcitiesto69.3%insizeVIcities.

InIndiatheshareofwalktripsincitiesgenerallyvarybetween20to30%acrossvaryingcitysizes.Ahighershareisobservedinlowerordercitieswhichkeepsdecliningascitiesgrowinsize.

Intermsofwalktripsitisobservedfromananalysisof30citiesofvaryingsizesthatthewalktripssharevarybetween16percentto34percentwithanexceptionof57%inhillyterraincitieslessthan0.5millionpopulationsize.Generallytheshareofwalktripstendtodecreasewithincreasingcitysize.Itwasalsoobservedthatsmallcitiesandparticularlyinhillyterrainexhibitaveryhighshareofwalk

tripsonaccountofcompactsize,densecore,greateraccessibilityandlessernumberofmotorizedvehiclesmakingthesecitiesimmenselywalkable.Itwasalsoobservedthatwalktripsaccountfor21.5%shareinbusinesstrips,10.5%forserviceand58%foreducationtravel.Basedondatapertainingtowalkabilityin30citiesitwasobservedthatwalkingispredominantmodeforurbanpoor.Itisfurtherseen that as the pedestrian environment continues to deteriorate the resident population is forced to accepthostileandunsafewalkingconditions.Thepaperwillalsohighlighttheanalysisofwalktrips.

Understanding social behaviors of pedestrian crowds in airborne images based on a graph-cut and sub-graph matching based framework

Beril Sirmacek, Deutsches Zentrum für Luft- und Raumfahrt, Oberpfaffenhofen GERMANY

Automaticunderstandingofhumanbehaviorsbecameaveryattractivetopic,sincethesesystemscanprovide crucial information to police departments or to crises management teams in order to prevent disastersorotherunpleasantsituations.Therefore,recentlymanyresearchersslidtheirfocusintothistopic.Inthelastdecade,manystudieshavebeenproposedtodetectandtrackpeoplefromgroundcameraimageswhicharealsoknownasclose-rangeimages.Mostofthesestudiesfocusedondetectingboundariesoflargegroupsandextractingglobalattributesofthem.Unfortunately,streetthesecamerashavelimitedcoverageareatomonitorlargeoutdoorevents.Inadditiontothat,inmostofthecases,itisnotpossibletoobtainclose-rangeimagesorvideostreamsintheplacewhereaneventoccurs.Therefore,inordertounderstandbehaviorsofpeopleinverybigout-doorevents,thebestwayistouseairborne images which began to give more information to researchers with the development of sensor technology.

Inourpreviousstudies,wehaveproposednovelapproachestodetectandtrackpeoplefromairborneimages.Inourfirststep,wehavefocusedondetectionofdensecrowdboundariesfromairborneimages[1].Byimplementingaself-parameter-adaptationmodule,wecouldalsodetectdensecrowdboundariesindifferentscaleinputimagessuchashighresolutionsatelliteimages[2].Infollowingstudy[3],wehaveextendedthestudyandbesidesdetectingdensecrowdboundarieswecouldalsodetectlocationsofindividualpersonsinsparseregions.Usingregisteredairborneimagesequences,wetrackeddetectedindividualsbasedonKalmanfilteringprocess.WithButenuthetal.[4],wehaveworkedonunderstandinghumanbehaviorsusingmotiontrajectoriesofindividuals.Althoughobtainedresultswerepromising,obtainingmotiontrajectorieswasnotimplementedasanautomaticprocess.Thereforewecanconcludethatautomaticallyunderstandinghumanbehaviorsisstillanopenandchallengingproblem.Inordertobringanautomaticsolutiontothisproblem,hereinweproposeanovelmathematicalapproachbasedongraph-cutandsub-graphmatchingprocesses.WestartwithgeneratingagraphnetworkforeachframeoftheinputsequencebyassumingdetectedpersonlocationsasnodesofthegraphandtheEuclideandistancebetweenthemasgraphedges.Thenwecutobtainedgraphintosub-graphsbycheckingtheirneighborhoodnodeproperties.Foreachobtainedsub-graphwegenerateafeaturevectoragainusinggraphpropertiessuchas;meannumberofneighborsforeachnode,sizesofsub-graph,velocityofsub-graphinpreviousframes,amountofnodenumberchangecomparingwiththepreviousframe.Aftergeneratingthedescriptorfeaturevectorforeachsub-graph,wecomparethemwiththedescriptorfeaturevectorswhicharepreviouslygeneratedandstoredinadatabaseforsamplesequenceslikemeetinggroups,diverginggroups,queuinggroupetc.Weapplyourmathematicalanalysisontoysequences.Additionally,wealsotestouralgorithmonanairborneimagesequence.Obtainedautomaticresultsindicatepossibleusageoftheproposedapproachonreal-scenesequences.

REFERENCES

[1] B. Sirmacek, P. Reinartz, „Automatic Crowd Analysis from Airborne Images“, 5th International Conference on Recent Advances in Space Technologies RAST 2011, Istanbul, Turkey, June 2011.

[2] B. Sirmacek, P. Reinartz, „Automatic crowd analysis from very high resolution satellite images“, Proceedings of Photogrammetry and Image Analysis Conference (PIA‘11), Munich, Germany, October 2011.

[3] B. Sirmacek, P. Reinartz, „Kalman Filter Based Feature Analysis for Tracking People from Airborne Images“, ISPRS Workshop High-Resolution Earth Imaging for Geospatial Information, Hannover, Germany, June 2011.

[4] M. Butenuth, F. Burkert, A. Kneidl, A. Borrmann,F. Schmidt,S. Hinz,B. Sirmacek, D. Hartmann, „Integrating pedestrian simulation, tracking and event detection for crowd analysis“, Proceedings of the 1st IEEE ICCV Workshop on Modeling, Simulation and Visual Analysis of Large Crowds, Barcelona, Spain, November 2011.

Evacuation analysis for venues: Systematical approach and comparison to evacuation trials

Sven Hebben, TraffGo HT GmbH, Flensburg GERMANYPatrick Gessler, TraffGo HT GmbH, Flensburg GERMANYHubert Klüpfel, TraffGo HT GmbH, Flensburg GERMANY

SCOPEOFWORKThispaperillustratesthesystematicalapproachforevacuationsimulationsandthecomparisonoftheresults to real world data for calibration and validation.

EvacuationAnalysesareatoolusedinFireSafetySciencetoassessbuildinglayoutsandvenuelocations.Oneareaofapplicationistooptimizepersonflows.Anotheroneistoassessdeviationsfrompresciptiveguidelines,e.g.concerningthedimensioningofescaperoutes.Forcertaintypesofbuildings,prescriptiveguidelinescannotbeapplieddirectly.Thismightbethecaseforarchitecturallydemandingbuildings,forsportsstadia,etc.Inthatcontext,problemsofdifferentproceduresordefinitionsofacceptancecriteriamightarise.ThereasonisthatprescriptiveguidelinesforbuildingsinGermanyandAustriadonotprovideanycriteriaconcerningevacuationsimulations,evacuationtimesor maximum densities for escape routes.

InGermanytherearetwoprojectstodevelopcertainboundaryconditionstostandardizetheprocedureofevacuationanalyses.OneoftheseistheRiMEAproject(RichtliniefürMikroskopischeEntfluchtungsanalysen-translated:GuidelineforMicroscopicEvacuationAnalyses),whichdefinestherequiredaspectstoconsiderwhilepreparingandanalyzingabuildingproject.Thesecondoneischapter9ofthevfdbLeitfadenIngenieurmethodenimBrandschutz(vfdbGuidelineforengeneeringmethodsinfiresafety).Thisguidelinesspecifiessomebasicparametersfortheinputofevacuationcalculations.

Inthispaper,twoexamplesofapplicationforthatprocedure(i.e.analysisaccordingtoRiMEAincombinationwithvfdb)areshown.OneisaentertainmentshowintheGrazcityhallinAustria,theotheroneisaTVstudioinGermany.

EVACUATIONTRAILSThefirstoftheevacuationexerciseswasconductedinGrazinAugust2010.ThisexercisewaspreparedbythefireBrigadeofGrazintheCityHall.ThevenueinwhichtheexercisetookplacewastheupperflooroftheexhibitionhallA.Theroomhadacapacityofabout1250persons,whichwereplacedinseatrowsinfrontofastage.Theevacuationshouldtakeplaceintheeveningduringaprogramofdifferentpresentationsandapoliticalcabaret.Thepersonsparticipatingwereinformedabouttheevacuationexerciseinthebeginning.Thepointintime,inwhichtheevacuationwasstarted,hadbeenconcealedfrom the persons present.

Thehazardscenariowascharacterizedbyabombthreatwithafollowingignitionintheentrancehall.Tocreateamorerealisticsituation,fogmachinesandsoundsystemswereusedtosupportthescenario.Duetothisscenario,themainentryofhallwasclosed.Thepersonswereinstructedtoleavethebuildingviatheemergencyexits.Theevacuationwasassistedbymarshals,whichwereplacedintheroom during the bomb threat.

ThesecondexerciseinaTVStudioinColognewasconductedtwice.Inbothexercisesabout450personswereseatedongrandstands.TheywereinstructedtoleavethestudioviatheemergencyexitsshortlyaftertherecordingofaTVshow.Theexerciseswereperformedwithdifferentconditions.Thefirstonewasannounced,thesecondonewasunannounced.

SIMULATIONThemodelusedtoperformtheevacuationanalysesisPedGo.Thebasisofthissimulationmodelisamultiagentmodel,whichisbasedonacellularautomaton.Theagentsarerepresentedindividuallyandtheircharacteristicsaredefinedbyindividualparameters.Thosearewalkingspeed,reactiontime,dawdle,sway,inertia,andgroupbehaviours.TheparametersarechosenaccordingtothespecificationsgiveninRiMEA.Thegeometryofthebuildings,whicharedefinedbyCADdrawings,aredividedintoacellulargridwithacellsizeof0.4x0.4m.Theresultsofthesimulationsshownandcomparedtotheresults of the evacuation trials for calibration and validation purposes.

Motions effect for crowd modeling aboard ships

Konstantinos Kostas, Technological Educational Institute of Athens, Athens GREECEAlexandos Ginnis, National Technical University of Athens NTUA, Athens GREECEConstantinos Politis, Technological Educational Institute of Athens, Athens GREECEPanagiotis Kaklis, National Technical University of Athens NTUA, Athens GREECE

VirtualEnvironmentforLifeOnShips(VELOS)isamulti-userVirtualReality(VR)systemthatsupports designers to assess (early in the design process) passenger and crew activities on a ship for bothnormalandhecticconditionsofoperationsandtoimprovetheshipdesignaccordingly[1].ThecrowdmodelingcomponentofVELOSisbuilduponthesteeringbehaviorstechnologyandrelatedenhancementsthatallowforconsiderationofpassengergroupingandcrewassistancebehavioreffectsinshipevacuationsimulations[2].

Furthermore,VELOSprovidescommunicationinterfacesenablingdataimportfromcomputationalpackages,includingsea-keepingandfireeventsmodelingsoftware.Pre-computedship-motionhistoryhasbeenusedinVELOSasasimplemeansforconsideringtheeffectofshipmotiononsimulatedpassengers’ movement via the introduction of the inclination steering behavior. Inclination behavior resemblesindefinitionandeffecttheinfluenceofagravityfieldthatwouldhinderagentmotionaccordingly.Theaforementionedapproachisasimplekinematicmodelthatdoesnotaccountforthedynamicnatureofthephenomenonthusignoringmotionaccelerations.Ship-motionaccelerations,however,arecriticaltotheassessmentofaperson’sbalancingand/orslidingaboardshipsandconsequentlytoitscapabilityofperforminganassignedtask.Inthiswork,wearefocusingontheexploitationofpre-computedshipmotionaccelerations,readilyavailablebytheconnectedsea-keepingcomputationalpackages.BasedontheworksbyGraham[3]andCrosslandetal[4]weinvestigatetheusageofMotion-InducedInterruptions(MIIs)andtippingcoefficientsinmodelingtheeffectofship-motion accelerations aboard.

Specifically,inaccordancetothecalculatedMIIrisklevel,wemodifysteeringbehaviors’weightingand/orbehaviorparameterstoaddressthecorrespondingpossibilityofamotioninterruptionanddegradationofassignedtasks’effectiveness.Theassumptionemployedhereisthataship-motion-interruptedsimpletask,suchaswalkingfrompointAtopointBwithMIIincidents,couldequivalentlybemodeledbyamodificationofthesteeringbehaviorsblendingthatwouldleadtosimilartaskeffectiveness.Thisbecomesmoreevidentifweconsiderthefollowingbasicsteeringbehaviors:ObstacleAvoidance,WanderandSeparation.AspersonsarefacedwithseastatesthatleadtohighpossibilitiesofMIIincidentstheirobstacleavoidanceandseparationcapabilitiesaredegraded,whilethewanderingcomponentreceivesamoresignificantcontribution.Thismodelingapproachiseffectivewithincertainbounds:alowerthresholdthatsignifiesthebeginningofnon-negligibleshipmotioneffects,andanupperthreshold,beyondwhichtaskcompletionisimpossible.

Ourapproachisfinallydemonstratedandanalyzedfortwotest-cases:A)anormalconditionscenariowheretypicalpassengerandcrewmovementsaresimulatedandB)ahecticconditionscenariowhere

anevacuationsimulationforaship’saftverticalzoneisperformed.Bothscenariosareexaminedwithandwithoutconsiderationoftheshipmotionseffectonsimulatedpassengerandcrewmovement.

REFERENCES

[1] A.I. Ginnis, K.V. Kostas, C.G. Politis, & P.D. Kaklis, 2010, “VELOS: A VR Platform for Ship-Evacuation Analysis”, JCAD special issue on Computer Aided Ship Design, volume 42, issue 11, p.1045-1058.

[2] K.V. Kostas, A.A-I Ginnis, C.G. Politis & P.D. Kaklis, 2011, “Use of VELOS platform for modelling and accessing crew assistance and passenger grouping in ship-evacuation analysis” in IMAM 2011 Genoa proceedings : “Sustainable Maritime Transportation and Exploitation of Sea Resources”, Eds E. Rizzuto, C. Guedes Soares, vol. 2, pp. 729-736.

[3] R. Graham, 1990, “Motion-Induced Interruptions as Ship Operability Criteria”, Journal of Naval Engineers, vol. 102, issue 2, pp. 65-71.

[4] P. Crossland, M.J. Evans, D. Girst, M. Lowten, H. Jones and R.S. Bridger, 2007, “Motion-induced interruptions aboard ship: Model development and application to ship design”, Journal of Occupational Ergonomics, vol. 7, pp. 183-199.

Early-stage egress simulation for process-driven buildings

Gabriel Wurzer, AIT Austrian Institute of Technology, Wien AUSTRIA

ABSTRACTManycomplexbuildingssuchashospitals,airportsandindustrialfacilitiesareprocess-driven,meaningthattheirdesignisconceivedaroundthedailyworkroutinesofthestaff,usuallycapturedusingbusinessprocesses(e.g.byusingflowchartsorBusinessProcessModelingNotation).Themainidea and contribution of our approach is to leverage such a static process model in order to facilitate adynamicegresssimulation.Indetail,weperformaprocesssimulationuntilaspecifiedtimetisreached.Asresult,wegetthetypicallocationoftheworkingstaffaswellasoccupancyofallareasofthebuilding,whichwethenfeedintoanagent-basedegresssimulation.Asresult,wecanobtaintheevacuationperformanceattimetunderconsiderationofthebuilding’sprocessmodel,i.e.differentusagescenariosthroughouttheday.Thishybridapproachbetweenprocesssimulationandpedestriansimulationisespeciallysuitedforearlystagesofbuildingdesign,whendifferentspatialconfigurationsandprocessvariantsareunderconsideration.Inthiscontext,theapproachisjustonepartofmanylinesofarchitectonicalreasoning,coveringforemosttheproblemofreachabilityandadjacencybymeans of pedestrian simulation.

ELABORATIONTheplanningofprocess-drivenbuildingsisoccupiedforemostwiththeproductionofaspatialdesignthatfacilitatesthedailyworkroutinesofthebuildingusers.Especiallyintheearlyphasesofthedesignproject,architectsandorganizationplannersworkinclosecooperationtoachievethisgoal:

• Organizationplannersdefinetheoperationalmodeloftheorganization,whichtakesthe formofahierarchyofbusinessprocesses(processesandsub-processes),anorganizational schemacontainingtheresponsibilitiesforeachprocess(i.e.processroles),and,ultimately,a definitionofstaffingneedsarisingoutoftheprocessmodel.• Architectsdefineaspatialconfigurationthatsatisfiestheorganizationalrequirements,most prominently:thebusinessprocesses.Inthiscontext,multiplevariantsofapreliminarydesign (alsocalledarchitectonicalschema)arebeingevaluated,thekeycriterionbeingtheadjacency of areas which exhibit a high degree of cooperation (as described by the process model).

Inpreviouswork[1],processsimulationhasbeenusedtosuperimposebusinessprocessesoverthepreliminarydesign,inordertoconstrainthespatialconceptsothatitfulfillstheprocessmodelandatthesametimevisualizeworkroutinesofthestaff.Thispaperextendsontheseconcepts,byintroducinga (dynamic) egress simulation that is based on the state of the aforementioned (static) process simulationatatimet.Morespecifically,we:

1. Thestateoftheprocesssimulationattimetgivestheusageofeachspace.2. Anegresssimulationtakesthisusageandcomputesthepedestrianflowtothenearestexits.3. Therecordeddensityandevacuationtimescanthenbevisualizedandsubsequentlyusedto getaninsightintobottlenecksoccurringbecauseofthespatialarrangementandexpected occupationofthebuildingattimet,atquiteanearlystageintheproject.

SUMMARYWeproposeahybridapproachlinkingstaticprocesssimulationtodynamicegresssimulation,inordertovisualizebottlenecksandreachabilityproblemsforprocess-drivenbuildings.Incontrasttoexistingegresssimulations,oureffortsaretargetedatearlystagesofdesignandactasqualitativemeansratherthangivingquantitativestatements.Thechoiceofprocesssimulationandpedestrianalgorithmisfree(theauthorsuseanownprocesssimulationlinkedtoaBlue-Adler[2]model).

REFERENCES

[1] Wurzer, 2010: „Schematic Systems -Constraining Functions Through Processes (and Vice Versa)“, In International Journal of Architectural Computing, Vol. 08 (2010).

[2] Blue and Adler, 2000: „Cellular automata model of emergent collaborative bidirectional pedestrian dynamics“, In Proceedings of Artificial Life VII, Portland.

Experimental study of the passengers’ self-learning in spatial structure of urban mass transit stations with cognitive map

Xiaolei Zou, Tongji University, Shanghai CHINAXiaoyi Qu, Shanghai Open University, Shanghai CHINAHaibo, Zhang, Hangzhou metro Corporation Ltd., Hangzhou CHINAJia Gao, Tongji University, Shanghai CHINA

Pathfinding,routingandpathplanningareimportantresearchfieldsinpedestriancrowdsimulation.Theexistingalgorithmsaremostlybasedonthethoughtsofoptimizationandpreferthepathswithshorterdistance,lesstime,andlowercost.Soitishardtoreflecttherealtrafficassignmentandmovingpathsofthecrowdwithasinglepathalgorithm.Andmanycrowdsimulationcasesconductedoverlyoptimistic results in crowd control and evacuation because they ignored the variations in individuals’ preference and familiarity to the environment.

Inthispaper,passengersofurbanmasstransitstationarestudiedontheirchanginginspatialcognitionandpathpreferenceovertime.Andasimulationframeworkisdefinedbasedonaself-learningprocessandthediversityofpathfindingandroutingoftheindividualstomakethecrowdsimulationmorerealistic.

THEUSEOFCOGNITIVEMAPSCognitivemapisgeneratedbytheexperiencesofpeople.Itisatoolforpeopletocollect,organize,storeandprocessthespatialinformation.Inthispaper,itisusedintheexperimentsandanalysis.Itstorestheinformationofspatialstructureofthestation,thelayoutofserviceequipment,theserviceprocedures,assignmentofpassengersandusedpathsforeachexamineeconstantly.

RESEARCHMETHODESANDPROCESSTheexperimentswereconductedattheinterchangestationofPeoples’squareandacommonstationofmetrolineNo.10inShanghai.100freshmanstudentswhowerenewtothestationswereselectedaspassengers.Theyweredividedintofourgroupstorepeat10timestheproceduresofalighting,departureandinterchangingatthe2stationsanddrewtheircognitivemapsaftereachexperience.Andthegroupswererequiredtorepeattheprocedureswithdifferenttimeinterval.Inthisway,thevariationsinsizeandfunctionofthestation,intrippurposeofpassengers,andinpracticefrequencywereconsidered.Thelearningcurveindifferentexperimentconditionscouldbedrewandcompared.

CONCLUSIONANDINSPIRATIONI.TheconstraintsforpathfindingandroutingAccordingtotheexperimentandsurvey,themainconstraintsforpathfindingandroutingarethedistance,thepassengerservicelevel,thedensityofthecrowd,andthecomfortlevelofwalkingalongthepath.Theseconstraintscomposethepreferenceofthepassengersinpathfindingandrouting.

II.ThespatiallearningcurveandchangingfrompathfindingtoroutingTheconstraintsabovebecomeclearerastheindividualspracticeandlearnmoreaboutthestation.

Thecognitivemapbecomesmorecompleteandthepathfindingturnstoroutingandthechoiceoftheindividual becomes more rational.

III.TheroutingandassignmentofthecrowdAtanytime,thecrowdcontainsindividualswithdifferentcognitivemapsofthestation.Thevariationsinpreferenceandfamiliaritylevelmakeadiversifiedroutingoftheindividualsandassignmentofthecrowd.

IV.TheSimulationFrameworkBasedonCognitiveMapandSelf-learningProcessThepaperproposesacrowdsimulationframeworkforurbanmasstransitstation.Theframeworkcontainsalibraryofbothpathfindingandroutingmodelsthatsuitfordifferentpreferencesandasetofcognitivemapsforindividualsindifferentself-learningstage.Itbecomespossibletoconstructmorerealistic simulation environments for representing the movement of passenger crowd in stations for operationoptimizationandalternativeevaluationforsafetyandefficiencypurpose.

REFERENCES

[1] Peng Gao, Ruihua Xu and Xiaolei Zou. 3-tier Architecture for Pedestrian Agent in Crowd Simulation. Pedestrian Evacuation Dynamics 2008. Springer, 2010:585-596.

[2] WAN Gang, GAO Jun, LIU Ying-zhen. Research on Cognitive Map Formation Based on Reading Experiments. Journal of Remote Sensing. Vo.12,No.2 Mar., 2008. P339-346.

Agent-based simulations of pedestrian movement for site security: U. S. Secret Service‘s current capabilities and next steps

Douglas Samuelson, InfoLogix Inc., Annandale VA USA

FortheU.S.SecretService,weassessedthestateofcurrentcapabilitiesinagent-basedsimulationsofmassegressandevacuation.Weconsiderhowtoaddresssomeofthemostsignificantchallengesatpresentindevelopingandusingsuchmodels.Wefocusinparticularonbehaviorslikelytooccurincrisisegresssituationsandtheeffectsthosebehaviorsarelikelytohaveonmanagingegress.Ofthese,themostsignificantare:

- Realisticvariationinspeedsofmovement,includingtheeffectofdifferentmixesof demographic characteristics of people in the crowd. - Variationsintheeffectofthesituation:obscuredsightlines,interferencewithvision,signage, apparent urgency.- Groupmovementsandtheconsequentdelays.- Counter-flowsinsomesituations,suchasparentsgoingbacktoassistchildren.- Movementbydisabledpeopleandotherstryingtoassistthem.- Effectsofdirectionandinstructionandofvariousways,bothatthetimeinadvance,of conveying this information.- Effectsofknownorsuspectedsecondaryhazards,suchastoxicplumes.- Effectsofplacementofresourcesandcommandposts.- Contra-flow,theingressofemergencyresponders.- Pre-positioningresources,suchasfirefightingequipment,medicalaidstationsandpersonnel.- Effectsofalternativeprovisionsformedicalcareon-siteorkeyedtotransportationoff-site.

TheU.S.SecretServiceaddressedthecontinuingissueofspeedofexecutionversusfidelityanddetailbydividingitsefforts:alarge-scaleplanningmodel(EvacuationPlanningTool,orEPT)andasmaller-scale,morerealisticSiteSecurityPlanningTool(SSPT)fortrainingprotectivepersonnel.SSPTisvideo-gamebasedandhigh;yadaptable,wellsuitedtoitspurpose.TheSecretServicealsoexploredfederating(linking)thesemodelswiththeU.S.NavelResearchLaboratory’sCT-Analystmodeloftoxicplume dispersion.

Puremodelingissuesstillposesignificantchallenges.The“HolyGrail”ofmodelingmassegressistohaveasingleuserinterfacethroughwhichonecanaccessrealisticlarge-scalemodelsofvenue-scaleandneighborhood-scaleevents,runninginnearrealtime,forexercises,planning,andassessmentofalternativecoursesofaction;smaller,highly-realisticreal-timemodelsforuseintrainingsecuritypersonnel,andsituationawarenessreal-timepresentationsforuseinactualincidentmanagement.Thisstructurewouldmakeitpossibleforincidentmanagerstotrainandplanwiththesamedepictionstheythenseeinactualoccurrences.Thisseamlesslyintegratedcapabilityisstillseveralyearsawayatbest,

however,sosomecurrenttrade-offsandadvancementsareindicated:

- Fasteroverviewmodelsofincidents,possiblybycompromis-ingdetailandrealismin depiction. - Largerandmorecomprehensivemodelsofsitesecurityfortraining.- Faster,morerealisticsmaller-scalemodelsofsitesecurity,runninglikevideogames,for specifictrainingexercises.- Moreattentiontoeventualconnectionsanddockingamongmodels,includingthedesign considerationsinvolvedinmakingscale,nomenclatureandunitsofmeasurementcompatible.- Makingdigitaloutputaswellasgraphicsstandard,sothatthenextmodelcanusethedigital version of the previous model’s computations as input for its own calculations.- Moreattentiontoeaseofmodificationofmodels.- Easierandmorewidespreadaccessbylocallawenforcementandeventsecuritypersonnel.

Therapidproliferationofeffortsinthisareaandtheacceleratingadvanceoftherelevanttechnologymakesflexibilityofdesigncriticalifmodelscurrentlyunderdevelopmentaretoremainusefullongenoughforincidentmanagerstolearn,train,analyzeandworkwiththem.SecretService’srecentandcurrent projects represent a substantial step forward in addressing this complicated and vital issue.

Comparison of simulation results with evacuation test and normative calculation

Martin Lopusniak, Technical University of Košice, Košice SLOVAKIA

Designofescaperoutesiscarriedoutaccordingtoastandardprocedureusingsimplemathematicalformulas.Thisprocedureisverylimitedforbothanarchitectandanownerofabuilding.Accordingtothisprocedureisnotpossibletocarryoutanydepthanalysisofanevacuation.Theresultisasimpletimeofapeopleevacuationfromabuilding.Theuseofevacuationmodelsgivesauseranabilitytocarryoutacomprehensiveanalysisofapeopleevacuationfromabuilding.Theresultisnotonlythetimeofevacuation,butalsoacomprehensivequantitativeandqualitativeanalysisofapeopleevacuation.VerificationofanevacuationmodelisanessentialrequirementforitsuseinthelegislativeframeworkoftheSlovakrepublic.PaperpresentsaverificationofthereliabilityoftheevacuationmodelbuildingExodusaccordingtotheevacuationtestresultsaswellasagainsttheresultsfromthestandard calculation procedure.

ThebasicmethodforworkisthecomputationalsimulationusingtheevacuationmodelbuildingExodus.AsthenormativecalculationmethodwasusedthecalculationaccordingtoSTN920102-Evacuationofpeople.Theresultsofcalculationsweremutualcompared.Evacuationtestwasconducted in an university environment. In the test was created situation as to evacuate students from classestotheoutside.Atotalofsixrunswereperformed.Duringthetestwererecordedevacuationtimesaswellasdifferentphysicalmovementcharacteristicsoftestparticipants.Dataobtainedwerethenusedtocomparesimulationresultswiththeresultsoftheevacuationtest.Obtainedevacuationtimesfromthesimulation,normativecalculationandevacuationtestswascomparedinfiveseries,whicharelabeledS1toS5.Eachseriescontainsasetofresults,theextentofwhichdependsonanumberofvariationsininputdatacalculations.Theywerecomparedtoruns1to4withthesimulation-seriesS1.SeriesS2isacomparisonofruns5and6withthesimulationtest.Comparisonoftheresults from normative calculation with the results from evacuation test was also performed separately forruns1to4–seriesS3andruns5,6-seriesS4.MutualcomparisonofthesimulationandthenormativecalculationisaseriesS5.Physicalcharacteristicsofparticipantswereaccuratelyenteredintothesimulationandcalculationaccordingtoinformationobtainedbyquestionnaire.Populationandpopulation panels have been established in accordance with the test groups during the evacuation test. ThegeometryofspacewascreatedwiththeCADimport.Reactiontimeineachsimulationwasequaltozero.Nootherbehavioralmodificationsweremade.

TheachievedresultsconfirmthattheevacuationmodelcanbeclassifiedasreliableforuseinthelegislativeframeworkoftheSlovakrepublic.Whencomparingtheresultsoftheevacuationtestandthesimulationwasconsensusintheoverallevacuationtimefrom85to95%.Thelargestdifferencesand the smallest compliance of the results were on evacuation routs from each room. In this case itwasashorttimeperiodandshortdistanceforpeopletoescape.Basedonthedatamaydrawthe

necessary conclusions on the use of evacuation model for a design and assessment of escape routes in thelegislativeframeworkoftheSlovakrepublic.Forwideruseofevacuationmodelsmustbecreatedconditions and evaluation criteria.

Merging of pedestrian queues: a mean field approach for CA models

Daniel Weber, Universität Duisburg-Essen, Duisburg GERMANYFlorian Knorr, Universität Duisburg-Essen, Duisburg GERMANYMichael Schreckenberg, Universität Duisburg-Essen, Duisburg GERMANY

Merging processes of particle streams are a crucial aspect in the modeling of various transport phenomena.Westudyanalyticallyandnumericallythemergingofseparatepedestrianqueuesfromtwonarrowcorridorsintoasingle-lanestream.Pedestriansaretreatedasmovingparticlesonalatticewithhardcoreexclusionandparallellatticedynamics.Thiscorrespondstothefloorfieldcellularautomatamodels(3)ofpedestriandynamicswherespaceisdiscretizedinquadraticcells.Eachcellcanbeeitheremptyoroccupiedbyexactlyonepedestrian.Duetotheparallelupdateschemeconflictsmayarise in the intersection area when two pedestrians try to enter the same cell.

TosolvetheseconflictsweuseanadditionalfrictionparameterwhichwasintroducedbyKirchneret.al.(1).Byvaryingthefrictionparameterthemergingprocessofpedestriansfromdifferentstreamscanbevariedfrom“cooperative”to“competitive”.Contrarytopreviouswork(2),weconsideraMooreneighborhoodatthejunctionsuchthatpedestriansfromonequeuecanchoosebetweentwotargetcellstoentertheexitstream(4).

Fortheanalyticanalysisweemployameanfieldapproximationtoderiveresultsfortheaverageoutflowindependenceofthefrictionparameterandinflowrates.DuetotheMooreneighborhoodthemergingareaofthejunctionisextendedtotwocells.Consequently,thereexistfourdifferentconfigurationsoftheintersection.Theresultingadditionalcorrelationsaretakenintoaccountviaatwo-clusterapproximation.

Eveninthissimplecasefiveparametersarerequiredforarealisticmodelingofthepedestrianmovement:1.Thehoppingprobability(beta1)intotheintersectionareafromstream1.2.Thehoppingprobability(beta2)intothefirstcelloftheintersectionareafromstream2.3.Thehoppingprobability(beta3)intothesecondcelloftheintersectionareafromstream2.4.Thehoppingprobability(alpha)forparticleswithinandoutoftheintersectionarea.5.Thefrictionparameter(mu)forsolvingconflictsasaconsequenceoftheparallellatticeupdate.

Thetime-evolutionoftheoccupationprobabilitiesofjunctioncellsisgovernedbyasetoflinearequations.Inthesteadystatetheseequationscanbesolvedanalytically.Formthestationarysolutiononecanderivetheoutflowfromthejunction.Wediscussthedependenceoftheoutflowforthevariousparameters.

Additionally,wealsoconsiderasemi-deterministiccasewherealpha=beta1=beta2+beta3=1.Withthischoiceofparametersthereisnodawdlingsothatparticleswillalwaysmoveaheadifthereisanemptycellinfrontofthem.(Onecanthinkofthiscaseasanevacuationscenario.)Thestochasticcomponent of the system is reduced to the friction parameter and the choice of the target cell of particlesfromstream2andleadstoaconsiderablesimplificationofthemasterequationgoverningthetime-evolution.

TheanalyticallyderivedfunctionaldependenceiscomparedtoMonteCarlosimulationsofthesystem.Preliminaryresultsshowgoodagreementbetweenthetwoapproaches.

REFERENCES

[1] A. Kirchner, K. Nishinari, A. Schadschneider: “Friction effects and clogging in a cellular automaton model for pedestrian dynamics”, PhysRevE.67.056122, 2003

[2] D. Yanagisawa, K. Nishinari: “Mean-field theory for pedestrian outflow through an exit”, PhysRevE.76.061117, 2007

[3] C. Burstedde, K. Klauck, A. Schadschneider, J. Zittartz: “ Simulation of pedestrian dynamics using a two-dimensional cellular automaton”. Physica A 295 (2001), 507-525

[4] T. Kretz, M. Schreckenberg: “Moore and more and symmetry”, In: Pedestrian and Evacuation Dynamics ‚05, Springer, 2006

Simulation model for vehicle and pedestrian interaction considering road crossing activities

Carlos Pretto, Federal University of Rio Grande do Sul, Porto Alegre BRAZILBruno Werberich, Federal University of Rio Grande do Sul, Porto Alegre BRAZILHelena Cybis, Federal University of Rio Grande do Sul, Porto Alegre BRAZIL

Thispaperpresentsthedevelopmentofpedestrianmodelcapabletorepresentsfromsimpletocomplexpedestrians and vehicles interactions.

Theemulationofpedestriansinthetrafficenvironmentisacomplexproblem.Inordertorepresenttherealmovementsofpedestrians,modelsshouldbeabletosimulateseveralprocesses,includingpathplanning,senseandavoidanceofobstacles,andinteractionwithotherpedestrians.Mostresearcheffortsonpedestrianmodelingfocusoncrowdsimulationorpedestrianmovementandinteractioninurbanenvironments(1),andtodealwiththecomplexityofpedestrianmovement,authorsoftenuseartificiallifeandcellularautomataapproaches(2,3).

Thesimulationofpedestrianbehaviorandvehicletraffichastraditionallyevolvedseparately,andtheassociationoftheseresearchfieldsisdifficulttofindintheliterature.Pedestrianfocusedmodelsfrequentlyignore,orprovidesimplifiedrepresentationsofvehiclemovements.Pedestriancrossingsarecrucialelementsinthetrafficsystems.Thedevelopmentofsoundmodelingrepresentationsofpedestriancrossingscancontributetoimprovingtheefficiencyandsafetyofthesefacilities(4,5).

Adesignofapedestrianmodelshouldbeadequatetoitspurpose.Itisnotpossibletocoverallactivitiesandpeculiaritiesofthepedestrians,especiallyinaroadcrossingenvironment.Theinclusionofunrepresentativevariablescanimpairthemodelsperformanceandrobustness.However,neglectingimportantaspectscouldmakethestudyweakandtheresultsunrealistic.

Itisimportanttolistdifferenttypesofpedestrian‘sanddriver‘sbehavioratthetimeofcrossing,classifyingthemaccordingtotheirimportanceandrelevanceinapedestrianmodel,regardingitsmodelingcomplexityandcomputationalcost.Accordingtothisclassification,itispossibletodevelopanewpedestrianmodelconsideringthemostimportantaspectsofthepedestrians‘roadcrossingactivity.

Theinteractionbetweenvehiclesandpedestriansmaypresentdifferentlevelsofcomplexity.Thesimpler,andmorefrequentlyobservedenvironment,consistsonsituationswherepedestriansusesidewalksandcrosstheroadonlyattrafficlightsorpre-definedcrossingpoints.Underthesecircumstances,thereisnolongitudinalinteractionbetweenpedestriansandvehicles.Itispossibletostratifypedestrians’andvehicles’flow,andtheinteractionbetweenthemcanbedefinedbystraightforwardrules.However,insomecircumstances,interactioncanachieveanincreasedlevelofcomplexity,generatingchaoticstateswhereitisnotpossibletoseparatetheflowofpedestriansandvehicles. Most models formulations are not capable of representing this chaotic environment.

REFERENCES

[1] Hao, J., Xu, W., Mao, T., Li, C., Xia, S., Wang, Z., A semantic environment model for crowd simulation in multilayered complex environment. Virtual Reality Software and Technology archive Proceedings of the 16th ACM Symposium on Virtual Reality Software and Technology. Kyoto, Japan, p. 191-198.

[2] Hamagami, T. and H. Hirata. Method of crowd simulation by using multiagent on cellular automata. International Conference on Intelligent Technology, IEEE, 2003.

[3] Ronald, N. and L., Sterling. Modeling pedestrian behavior using the BDI architecture. International Conference on Intelligent Agent Technology (IAT´05), IEEE, 2005.

[4] Chae, K.. Simulation of pedestrian-vehicle interactions at roundabouts. PhD Thesis, North Carolina State University, Raleigh, NC, United States of America, 2005.

[5] Chu, X., M. Guttenplan and M. Baltes. Why people cross where they do: The role of the street environment. Transportation Research Record, v. 1878, 2004, p. 3-10.

[6] Meschini, L. And G., Gentile (2009). Simulation car-pedestrian interactions during mass events with DTA models: The case of Vancouver Winter Olympic Games. International Conference, SIDT, Milan, 2009.

Poster exhibition

Unsorted list

Nirajan Shiwakoti, Monash University, Melbourne AUSTRALIA Understanding crowd panic at turning and intersection through model organisms

Daisuke Fukuda, Tokyo Institute of Technology, Tokyo JAPAN An econometric based pedestrian walking behaviour model implicitly considering strategic or tactical decisions

Stefan Sarstedt, Hochschule für angewandte Wissenschaften, Hamburg GERMANY WALK: A modular testbed for crowd evacuation simulation

Alexander Kamyshnikov, Siberian Federal University, Krasnoyarsk RUSSIA Computational complexity and parallel computing in discrete-continuous SIgMA.DC pedestrian movement model

Tuomo Rinne, VTT Technical Research Centre of Finland, Esbo FINLAND Bottlenecks in evacuation design considering both structural and human behavioural aspects: An experimental study

Cecile Appert-Rolland, University Paris-Sud, Paris FRANCE A macroscopic model for bidirectional pedestrian flow

Barbara Krausz, Fraunhofer-Institut IAIS, Sankt Augustin GERMANY Integrating lateral swaying of pedestrians into simulations

Toshiyuki Kaneda, Nagoya Institite of Technology, Nagoya JAPAN Evacuation agent simulation in underground shopping street adding floor field approach and its three dimensional expression

Understanding crowd panic at turning and intersection through model organisms

Nirajan Shiwakoti, Monash University, Melbourne AUSTRALIAMajid Sarvi, Monash University, Melbourne AUSTRALIACharitha Dias, Monash University, Melbourne AUSTRALIAMartin Burd, Monash University, Melbourne AUSTRALIA

Herding,flocking,schoolingandswarmingarecommonlyobservedgroupingbehaviorsofanimals(1).Activitiesattransitstations,sporteventsorreligiousgatheringsarecommonexamplesofcollectivebehaviorsinhumancommunitiesundernormal,everydaysituations.Underpanicconditions,suchaspredatorattack,animalstrytostayclosetoeachotherinordertoequalizetheriskandreducechancesofbeingcaughtbyapredator(1).Similarbehaviorscanbeobservedinhumansduringescapefromburningbuilding,theatres,stadiumsetc.(2).Thus,insightsfromnon-humanbiologicalsystemsmaybe used to understand the collective human behavior under emergency situations and to enhance the safetyofhumancrowdsasdescribedin(3,4).

Thelayoutoftheescapeareacouldhavesubstantialnegativeeffectsoncollectivecrowdflowandbehaviour,especiallyunderpanicconditions,wherelimitedtimeisavailabletoescapefrompotentialdangers.Althoughpreviousstudiesoncrowddisastershavehighlightedtheimportanceofconsideringmovementpatternsatturningpointsandatintersection(2),verylimitedqualitativeandquantitativestudieshaveaddressedthisphenomenon,particularlyunderpanicsituations.Onereasonforthismightbethelackofempiricaldatatovalidatethepredictionsfrommathematicalmodels(5).

Inthiswork,weusemodelorganismsapproachbyusingempiricaldatacollectedfrompanickingArgentineantstostudycrowdpanicatturningandintersection.Experimentswereconductedwiththeantsforthedifferentangledcorridor(0,30,45,60and90)aswellasatrightangledintersectionunderpanic situation.

Fromthedataanalysis,itwasobservedthattheflowdoesnotalwaysdecreaseproportionatelywiththeincreaseinturningangles.Itshowsthattheremaybeoptimalanglesformaximizingflowandthatflexibility on the choice of turning angles can have implications in situations when it is not possible to havestraightcorridorduetodesignand/orspacerestrictionsincaseofhumansituation.Similarly,fromintersectionexperiment,itwasobservedthatthemostdangeroussituationcouldbewhenonestreamofflowofindividualsisblockedbyanotherstreamofindividualsforconsiderabletimeresulting in disproportionate flow at the intersection.

Thestudydemonstratethattheanimalmodelssuchasantcoloniesmightallowempiricaltestingofhumanpedestrianmodelswhenhumansubjectscannoteasilyorethicallybeemployed,providingthemodelcanbescaledappropriately.Successfulpredictionofcollectivemovementinbothspecieswouldsuggestthatcommonunderlyingdynamicsgovernthebehaviourofself-drivenparticlesacrossawidesizerange.

REFERENCES

[1] Okubo, A. Dynamical Aspects of Animal Grouping: Swarms, Schools, Flocks and Herds. Advance in Biophysics, Vol. 22, 1986, pp. 1-94.

[2] Chertkoff, M.J. and Kushigian, H.R. Don’t Panic: The psychology of emergency egress and ingress. Praeger Publishers, USA, 1999

[3] Shiwakoti, N., M. Sarvi, G. Rose, and M. Burd, M. Enhancing the safety of pedestrians during emergency egress: Can we learn from biological entities? In Transportation Research Record: Journal of the Transportation Research Board, No. 2137, Transportation Research Board of National Academies, Washington, D.C., 2009, pp. 31–37.

[4] Shiwakoti N., M. Sarvi, G. Rose, and M. Burd. Animal dynamics based approach for modeling pedestrian crowd egress under panic conditions, Transportation Research Part B, Vol. 45, 2011, pp. 1433-1449

[5] Helbing, D., Farkas, I. and Vicsek, T. Simulating dynamical features of escape panic.Nature, No. 407, pp. 487-490, 2000.

An econometric based pedestrian walking behaviour model implicitly considering strategic or tactical decisions

Daisuke Fukuda, Tokyo Institute of Technology, Tokyo JAPANToru Seo, Tokyo Institute of Technology, Tokyo JAPANKaoru Yamada, Tokyo Institute of Technology, Tokyo JAPANHideki Yaginuma, Highway Planning Inc., Tokyo JAPANNobuhiro Matsuyama, Merrill Lynch Japan Securities, Tokyo JAPAN

Analysingpedestrianbehaviourisimportantforbetterdesignoffacility,evacuationplanningandsoon.Previousapproachesofmodellingpedestrianbehaviourcanbeclassifiedintotwodirections:(1)cell-basedspatialrepresentation(e.g.cellarautomatamodel)and(2)continuousspatialrepresentation(e.g.socialforcemodel).Asthethirddirection,econometric-basedpedestrianbehaviouranalysisbasedondiscretechoicemodel(e.g.Antoninietal.2006andRobinetal.2009)hasbeenrecentlyattractingattentiondue.Thisapproachisattractivebecausevariousinfluentialfactorsforwalkingbehaviourcanbe easily incorporated into utility functions and also the calibration of model parameters is rigorous based on econometric theory.

Exitingeconometricbasedpedestrianmodels,however,onlyconsiderwalkingbehaviouratoperationallevel and do not deal with any choice decision at strategic level (e.g. destination) or at tactical level (e.g.route).Usuallyitisassumedthatstrategicandtacticaldecisionsareexogenouslymade.Invariousrealsituations,however,pedestrianshavebeenintermittentlymakingadecisiononre-changingtheirdestinationsorroutes.Forexample,videoobservationattrainstationstellsthatpedestrianswhoaregoingthroughticketgatewouldtendtochangetheirtargetgateenroutewhenthestationplazaandthearea neighbouring the gate are so crowded.

Anotherissueontheuseofeconometricbasedpedestrianmodellingisthatincalibrationitiscostlytocollectpedestriantrajectorieswhichareusuallymanuallytrackedfromvideosequences.Tosomeextent,automationindatacollectionofpedestriantrajectoriesishighlyexpected.

Inthisstudy,weproposeaneconometricbasedpedestrians’walkingbehaviourmodelwhichimplicitlyconsiderstrategicortacticaldecisions.Inrealsituations,itisonlypossibletoobservepedestrians’trajectoriesandtheirfinalchoiceresultsondestination(e.g.theticketgatetheyactuallychosen)andthetargeteddestinationenroutearelatentandunobservable.Toreflectthisfactamodelofdynamiclatentplans(Choudhuryetal.2010)hasbeenextensivelyutilisedbyassumingdestinationchoiceasadecisionatplanlevelandwalkingatactionlevel.Thelatentplans(destination)bypedestriansmaydynamically change subject to environment and it further leads to the dynamical and structural change intheiractionchoices(walkingtrajectories).

TheproposedmodelisvalidatedwiththevideosequencesofpedestrianbehaviourwhichwasrecordedaroundaticketgateatacrowdedrailstationinTokyo.Forefficientdatacollection,ageneralisedstate-spacemodellingapproach(particlefiltermethod)isalsoappliedtoautomaticallytrackwalkingbehavioursfromvideosequences.Theparametercalibrationresultindicatesthattheproposedmodelfitstherealdatamuchbetterthanthesimplemodelofwalkingbehaviour.Finally,apedestrian

simulator is implemented and tested for evaluating pedestrian movements in a larger and more crowded station.

REFERENCES

Antonini, G., Bierlaire, M., Weber, M., 2006. Discrete choice models of pedestrian walking behavior. Transportation Research Part B 40 (8), 667–687.

Choudhury, C., Ben-Akiva, M., Abou-Zeid, M., 2010. Dynamic latent plan models. Journal of Choice Modelling 3 (2), 50–70.

Robin, T., Antonini, G., Bierlaire, M., Cruz, J., 2009. Specification, estimation and validation of a pedestrian walking behavior model. Transportation Research Part B: Methodological 43 (1), 36–56.

WALK: A modular testbed for crowd evacuation simulation

Stefan Sarstedt, Hochschule für angewandte Wissenschaften, Hamburg GERMANYIa Enukidze, Universität Hamburg, Hamburg GERMANYStefan Münchow, Hochschule für angewandte Wissenschaften, Hamburg GERMANY

Whenlargenumbersofpeoplegatherinpublicspacessuchasstadiums,railwaystations,shoppingcentresandconcerthallsthereisanincreasedriskofmassemergenceanddisasters.Criticalsituationscouldpossiblybepreventedwithappropriatemeanstoanticipatethem.WALKisamodulardesignedtestbedforcrowdevacuationsimulationdevelopedattheHamburgUniversityofAppliedSciences.Itisdesignedrespectingtherequirementsofcivildefencebyallowingthedefinitionandsimulationofversatilescenarioswithdifferingenvironmentstructureandcrowdattributes.Inordertosimulateindividualagentsandprovideadistributionofthesystem,WALKisrealizedasamulti-agentsimulation.

Systematicstudiesofhumanbehaviourincriticalsituationsandquantitativetheoriescapableofpredictingcrowddynamicsarerare.Itisdifficulttoreenactadisasterormasscrowdinginstudytowatchhumanpanicbehaviourclosely.Researchersdifferinopinionregardingtypicalbehaviourofindividualsincriticalsituations.However,insomeareastheyagree.Crowdingbehaviouroccursifresources that are necessary to survive or resources that are very popular are getting scarce.Indangeroussituationsfearplaysanimportantrole.Whenhumansfeelthreatened,anarchaicpartofthebrainreactswithflightorfightinstinct.Theperformanceandpowerofjudgementofthebraindecreases.Thiscouldpotentiallyresultinbottleneck-effectsatknownexitswhilealternativeemergencyexitsareoverlooked.Peoplewillbesquashedortrampledtodeath.Toachieveanappropriatesimulationofhumanswithagents,emotions,personalityandsocialdynamicshavetobeconsidered.

WALKisbasedonaflexiblecomponent-basedarchitecture.Ageographicalinformationsystem(GIS)provides information about the simulated environment. Information layers are used to distribute requiredinformation,e.g.thecurrentairtemperature,acrossthesimulation.Agentsusetheselayersasabasisfortheirperception.OneofthekeyfeaturesofWALKisthepossibilitytodefinescenarios,whichallowsafastandflexibleapplicationofthesimulationsystem,e.g.incaseofananticipatedterroristattack.Ascenariospecificationisacompletedescriptionofthesetting,whichincludesarchitecture,sizeandcompositionofthecrowdandtheoccurrenceofevents.

WALKagentswillbeequippedwithcomponentsforsimulatingpersonalitytraits,emotionandsocialinteractiontoshowbelievablehuman-likebehaviour.Amodularagentarchitecturewillallowthereplacementofsinglecomponents.Thus,acomparisonofdifferentemotion,personalityandsocialbehaviour models will be possible and will help to draw important conclusions about the simulation of human-likebehaviour.Consequently,WALKcanserveasaninter-discliplinarytestbedforscientistsfromdifferentdisciplineslikesocialsciences,psychologyandcomputerscience.

Inthecurrentstateofdevelopment,abasicsimulationplatformisimplemented.SimplescenarioscanbedefinedandtheGISprovidesallnecessaryinformationtosimulatethem.Inordertoshowthesimulationstate,WALKprovidesreal-time2Dand3Dvisualization.Inthenearfuture,aflexibleagentarchitecturetosimulatedifferentpsychologicalmodelswillbedeveloped.

Computational complexity and parallel computing in discrete-continuous SIgMA.DC pedestrian movement model

Alexander Kamyshnikov, Siberian Federal University, Krasnoyarsk RUSSIAEkaterina Kirik, Institute of Computational Modelling RAS, Novosibirsk RUSSIAAndrey Malyshev, Institute of Computational Modelling RAS, Novosibirsk RUSSIAEgor Popel, Siberian Federal University, Krasnoyarsk RUSSIA

Thepaperdealswithcomputationalaspectsofpedestriandynamicsmodelling.Thediscrete-continuousstochasticfloorfieldmodelSIgMA.DCisconsidered[1].Themodelsimulatesmovementofasinglepersonintheflow,takingintoaccountotherparticles,obstacles,strategies(theshortestpathandtheshortesttime),destination,inacontinuousspace.Numberofthepossibledirectionsq,wherepersonmaymoveateachtimestep,isamodelparameter.Movementequationisgiveninadirectformas a function of a previous position and local particle’s velocity.

Frompracticalapplicationsofpedestrianmovementmodelscontinuityofaspaceisthepros,andconsofcontinuousmodelsareproblemswithdiscretizationofthetasktorealizesimulationsoncomputer.TheothercontributionstothecomputationalcomplexityofthesimulationaregivenbythenumberNofpeopleandsquaresofcrowedplaces,whichmayreachseveraltensofthousands.

Thefollowingtasksarethemosttimeconsuming.ThemodeladoptsanideaofusingthestaticfloorfieldSfromCAapproach[2,3].Toavoidboundaryeffectaveryfinemesh(upto1cminsidesize)isused.AspecialalgorithmisusedtocalculatethefieldSandsaveitinRAM.

Pedestrianvelocityisdensitydependent,andinourmodelthisdependenceisusedindirectform.ToestimatelocaldensityforeachparticleineachstepanddirectionweproposetouseMonte-CarlomethodcombinedwithNadaray-Watson’sweighteddensityestimate.

Tosavecomputationaltimeandtosimulateevacuationon-lineorfaster(thatisveryimportantforcomprehensivepracticalapplications)aparallelizationofthecomputationalalgorithmisused.Comparisonofpossibleparallelimplementationofthemodel(bypersons,floors,slices)isdiscussed.AsoftwareimplementationofthemodelisrealizedusingtheprogramminglanguageC++,QtframeworkandMPICH2libraryforhighperformanceparallelcomputing.

Therewaspaidattentiontoprogramarchitecture.Thecalculationengineismadeassingleexecutablemodulethatextendstheapplicationofthemodel.Forexample,forKrasnoyarskschoolswedevelopasoftwaretool“3D-simulatoroffireevacuation”consistingofacomputationalkernelthatsendsdatatothe3D-visualizationprogram.Thissimulatoristoimproveandmaintainapupil’scompetenceinthefieldoffiresafetyusingthevisualgamelearningform.

REFERENCES

[1] Kirik, E., T. Yurgel‘yan, A. Malyshev. On discrete-continuous stochastic floor field pedestrian dynamics model SIgMA.DC // In the book “Emergency evacuation of people from buildings”, 2011. – pp. 155-161.

[2] Kirik, E., Yurgel‘yan, T., Krouglov, D. (2011) On realizing the shortest time strategy in a CA FF pedestrian dynamics model // Cybernetics and Systems, vol.42:01, 1-15.

[3] Schadschneider, A., A. Seyfried. Validation of CA models of pedestrian dynamics with fundamental diagrams // Cybernetics and Systems, Vol. 40 (5), 2009, 367–389.

Bottlenecks in evacuation design considering both structural and human behavioural aspects: An experimental study

Tuomo Rinne, VTT Technical Research Centre of Finland, Esbo FINLANDPeter Grönberg, VTT Technical Research Centre of Finland, Esbo FINLANDTerhi Kling, VTT Technical Research Centre of Finland, Esbo FINLANDTimo Korhonen, VTT Technical Research Centre of Finland, Esbo FINLAND

Thequalityofdatafromnormalevacuationdrillsdependstronglyonhowwelltheevacueesareinformedaboutthedrillbeforehand.Forinstance,pre-movementtimedistributionsmightleadtowrongconclusionsinwell-informedsituations.Inaddition,peoplemovementisdominatedbytheawarenessoffamiliarroutesandnoanomaloushumanbehaviourisobserved.Usualreasonforwell-informeddrillsisthatdrillsareusuallyheldinfrequentlysothereisafearthatsomethingmightgowrongifthedrillisnotwell-informed.Atthemoment,normalfiredrillsrevealrarelybottlenecksinpointofviewoffiresafetydesign,buttheyareusefulforbuildingsafetyorganisationsandstaffforknowingthegenerallevelofsafetyandrecognisingtheweakpoints.Hopefully,infuturethemodernsafetycultureandthewayofthinkingcouldleadtothesituationwherenormalfiredrillswouldoffermoredifferenttestscenarios.

Usuallydimensioningproblemsinevacuationsituationsrelatetobottlenecks,e.g.,doorgeometries,stairs,mergingflows,counterflows,andemergencyrescuepersonneloperations.Theseissuescanbedemonstratedthroughfullycontrolledexperimentalset-ups.VTTisacoordinatorinaproject“Evacuationandrescueoperationsinfireconditions”thatwillproduceexperimentaldatainfourdifferentscenariosandgeometries.Thefirstgeometryrelatestostairs,wherefatigue,firefighters’andnormalpeoplecounterflows,andfirefighters’rescueoperationsinstairsareexamined.Thesecondtrialwillbeperformedusingdifferentdoorgeometries,wheretheaimistomonitorhowpronepeoplearetoformanewqueuenearthedoorswhenselectingadoorway,howcertaingeometricalarrangementsincreasespecificflows,andwhatistheeffectofdooropeningforcesandanglestothespecificflows.Twootherscenariosareconductedincorridors,wheredensityofsmoke,differentlightningconditions,routeselectionsandcounterflowsareexaminedonbothgroupandindividuallevels.TheinstrumentationconsistsofnormalDV-cameras,RFID-tags,andsensorsmeasuringangularaccelerationanddooropeningforces,andphysicalstressofthefirefighters.Inaddition,questionnaireswill bring more detailed information to explain individual data and behavioural phenomena.

ExperimentswillbeperformedinMarch2012andthepersonsattendingthetestsarestudentsofFinnishemergencyservicescollegeandconscripts,intotal150persons.Mostofthetestpersonsaremales,aged20-30yearsold.Dependingonthenatureoftheexperiment,thegroupsizevariesfrom15to40persons.Inaddition,someofthetestswillbeconductedonanindividuallevel.Theresultswillbepublishedinthefinalreportthatisfreelyavailableoninternetintheendofyear2012.Thepreliminaryresults of the test series without throughout statistical analysis are presented in the poster.

A macroscopic model for bidirectional pedestrian flow

Cecile Appert-Rolland, University Paris-Sud, Paris FRANCEPierre Degond, IMT University Toulouse, Toulouse FRANCESebastien Motsch, University of Maryland, College Park MD USA

Thisposterpresentsamacroscopicmodelforpedestriandynamicsinacorridor(orinanyquasione-dimensionalsystem)[1].ThemodelisinspiredfromtheAw-Rasclemodelofcartraffic[2]withthemaindifferencethathere,atwo-directionalflowisconsidered:ineachpoint,twodensitiesaredefined,respectivelyforleftandrightgoingpedestrians.Thechallengeistoimplementamaximaldensityconstraint in order that the density remains bounded even under congestion.

Tomodelthedensityconstraint,weintroduceapressureterm,tomodeltheinteractionsbetweenpedestrians,thatdivergeswhenthedensityapproachesthemaximaldensity.Theintensityofthedivergence is controlled by a small parameter epsilon.Inthelimitwhereepsilontendstozero,thepressuretermbecomessingularatthemaximaldensity.Thenthesystemexhibitscoexistingphases,withdensitieseitherequaltothemaximaldensity(congestedphase)orlessthanthemaximaldensity(uncongestedphase).Thesephasesareseparatedby sharp interfaces.

Avariantofthemodel,forwhichallpedestrianshavethesamepreferredvelocity,allowstohavesimplercalculationsandtoillustrateandstudybothimplementations(finiteorvanishingepsilon)ofthe congestion constraint.

Thelateralextensionofthecorridorcanbetakenintoaccountthroughamulti-lanemodel,withappropriate lane changes.

Acharacteristicoftwo-waymodelsisthattheycanloosetheirhyperbolicityinsomecases.Actually,thiscouldbethecounterpartofphenomenaobservedinrealcrowds,namelytheinstabilityofhomogeneousflowstowardslane-formation,orevencrowdturbulenceasobservedatveryhighcrowddensities[3].

Thepressuretermappearinginourmodelcouldinprinciplebeobtainedfromexperiments.Itisalsopossiblewiththisformalismtohavedifferentcharacteristicsforeachpedestrianflow.Forexample,pedestriansheadingtowardsatrainplatformcouldbemorepushythanthosegoingbackwards.Thisremains to be explored analytically and numerically.

Acomparisonbetweenthesimulationsbasedonthismodelandanexperimentperformedinaringcorridor[4]iscurrentlyperformed[5].

REFERENCES

[1] C. Appert-Rolland, P. Degond and S. Motsch, „Two-way multi-lane traffic model for pedestrians in corridors“, Networks and Heterogeneous Media, vol. 6, p351-381 (2011).

[2] A. Aw and M. Rascle, „Resurrection of ``Second Order‘‘ Models of Traffic Flow and numerical simulation“, SIAM Journal on Applied Mathematics, vol. 60, p916-938 (2000).

[3] D. Helbing, A. Johansson and H. Zein Al-Abideen, „The Dynamics of Crowd Disasters: An Empirical Study“, Phys. Rev. E, vol. 75, p046109 (2007).

[4] M. Moussaid, E. Guillot, M. Moreau, J. Fehrenbach, O. Chabiron, S. Lemercier, J. Pettre, C. Appert-Rolland, P. Degond, and G. Theraulaz. Outbreak and breakdowns of collective intelligence in human crowds. submitted to PLoS Computational Biology, 2011.

[5] S. Motsch et al, in preparation.

Integrating lateral swaying of pedestrians into simulations

Barbara Krausz, Fraunhofer-Institut IAIS, Sankt Augustin GERMANYChristian Bauckhage, Fraunhofer-Institut IAIS, Sankt Augustin GERMANY

Simulationsofpedestrianstreamsplayanimportantrolewhenoptimizingevacuationroutesoridenfiyinghazardouslocationsinabuildingoreventlocationinordertopreventaccidents.Obviously,itisofvitalimportancethatsimulationsareasaccurateandrealisticaspossible.Consideringfutureresearchdirectionsthatintegratereal-timeinformationintosimulationsinordertoforeseehazardoussituations,preventaccidents,andrecommendevacuationstrategiesinreal-time,accuratepedestrianmodels are even more important.

Inordertoimproveandvalidatemodelsofpedestrianbehavior,insightsintohumanmotioncharacteristicsareneeded.Inparticular,thedynamicsinlocationsofhighpedestriandensityareofgreatinterest.Averycharacteristichumanmotionpatternislateralswaying.Fromcommonobservations,itisknownthatpeopledonotmovealongastraightline,butinsteadtendtoswinglaterally.Recently,thischaracteristicmotionpatternhasbeenexploitedinordertodetectcongestedareasbyanalyzingshort-termmotionsfromvideosurveillancecameras[3,4].

However,acomparisonofsimulatedtrajectoriesandrealtrajectoriesrevealsthatlateralswayinghasnotbeenadequatelytakenintoaccountinstate-of-the-artpedestrianmodels.Inthiswork,weextendthegeneralizedcentrifugalforcemodel[1,2]byintroducinganoscillationforcethatmimicslateralswaying of pedestrians.

First,weshowthatthereisalinearrelationshipbetweenthevelocityandtheamplitudeoflateralswayingaswellasbetweenthevelocityandthefrequency.Forthatpurpose,weanalyzerealtrajectoriesobtained from video recordings of a large scale experiment conducted under laboratory conditions. Next,inadditiontothedrivingforceandtherepulsiveforcesproposedinthegeneralizedcentrifugalforcemodel,wesuperimposeanoscillationforceusingtheobtainedparametersforfrequencyandamplitudeoflateralswaying.Inadditiontointroducingtheoscillationforce,wealsoadapttheellipsemodelingspacerequirementsofpedestrianstoareasonablesize.

Toquantitativelyevaluatetheeffectivenessofourmodel,wesimulatepedestrianmovementsandshowthatthefundamentaldiagramiswellreproduced.Secondly,weemploytrajectoriesobtainedfromalarge-scaleexperimentunderlaboratoryconditions,usetheirinitialpositionsandsimulatepedestrianmovements.Acomparisonofrealtrajectoriesandsimulatedtrajectoriesrevealsthatourmodelsimulates realistic trajectories.

REFERENCES

[1] M. Chraibi and A. Seyfried. Generalized centrifugal-force model for pedestrian dynamics. Phys. Rev. E, 82(4):046111, 2010.

[2] M. Chraibi, A. Seyfried, A. Schadschneider, and W. Mackens. Quantitative description of pedestrian dynamics with a force-based model. In Int. Conf. on Web Intelligence and Intelligent Agent Technologies, 2009.

[3] B. Krausz and C. Bauckhage. Automatic detection of dangerous motion behavior in human crowds. In AVSS, 2011.

[4] B. Krausz and C. Bauckhage. Loveparade 2010: Automatic video analysis of a crowd disaster. CVIU (to appear), 2011.

Evacuation agent simulation in underground shopping street adding floor field approach and its three dimensional expression

Toshiyuki Kaneda, Nagoya Institite of Technology, Nagoya JAPANYoshiyuki Kobayashi, Nagoya Institite of Technology, Nagoya JAPANMasaki Tamada, Kozo Keikaku Engineering Inc., Tokyo JAPANTaichi Shimaru, Kozo Keikaku Engineering Inc., Tokyo JAPANKeisuke Hata, Kozo Keikaku Engineering Inc., Tokyo JAPAN

InJapanundergroundshoppingstreetshavebeendevelopedinlargecities,wherecitizenshavebeenlovedascommutingpathsandshoppingspaces.However,ontheotherhand,undergroundshopping street as a built environment is exposed to vulnerability from the viewpoint of mainly air ventilationandlightingmatters,thussufficientconsiderationforspacedesignandevacuationstrategyarerequired,especiallyinthecasesoffireandearthquake.Inthiscontext,aftercomprehensivedecentralizationlawenforcedin2000,citygovernmentsareempoweredthediscretioninspacedesignandevacuationstrategyonundergroundstreets,soitisexpectedtobeassessedthroughrealisticsimulation.

ThispaperreportsatrialofevacuationsimulationforexaminingspacedesignandevacuationstrategyinanundergroundshoppingstreetindowntownNagoyainthecasesofelectriclightingblackout.Atthattime,weemploythree-dimensionalspaceexpressionintelligibleformanystakeholdersthrougheach process of the simulations.

TheauthorshavealreadydevelopedapedestrianagentsimulationmodelthatwecallASPF(AgentSimulationofPedestrianFlows).ASPFver.1startedasarule-basedagentmodelwith40cmx40cmcellsizeon2Dspaceandahalfsecondperstep.ASPFver.2analyzedAkashioverpasscrowdaccident;ASPFver.3realizedthatagentcanwalkona‚pseudo‘continuousspaceforalldirections,andASPFver.4examinedcomplicatedspacesaswalkingenvironmentsinthatwelettheimprovedagentswalkalongthelinesinnetworkofwaypoints.

Here,consideringevacuationincasesofthenormalsituationisonewayflowsinprinciple,whenwetrytosimulatenaturalisticevacuationflows,wethoughtthatanintroductionofthefloorfield(staticfield)potentialapproachhasadvantageinthepointsofconvenienceandpracticality,forreducingcomputationalcomplexity.SowenewlydevelopedASPFver.5.Thismodelisimplementedonanagent-modelingplatformartisoc3.0,andeachagentbehaveswith22rulesforotheragents,14rulesforwallslikeasver.4.InASPFver.5,differentfromBursttedeetal’sprobabilisticapproach,oneverystepeachagent updates one’s direction to the center of the floor cell of the minimum potential value among adjacent8cells(Mooreneighborhood),andwhentherearetwoormorecellsoftheminimum,thisagent doesn’t renew one’s direction.

Moreover,weevaluatedtheevacuationperformanceofeachspacethroughsimulationsincomparisonwith the results of the cases of the three space shape models based on the actual measurement survey of aundergroundshoppingstreetindowntownNagoya.Indiscussion,wealsorefertheexpandabilityofthefloorfieldapproachtowardthree-dimensionalspace.

Poster exhibition

Unsorted list

Charles Thornton, Thunderhead Engineering, Manhattan KS USA New wayfinding techniques in pathfinder and supporting research

Stavros Katsoulis, Village roadshow group of companies, Athens GREECE An innovative evacuation system for multiplex cinemas: Case Study “Village roadshow group of companies - Athens“

Henri Hakonen, KONE Corporation, Espoo FINLAND Simulation models of merging priorities in staircases

Wael Alhalabi, Umm AlQura University, Makkah SAUDI ARABIA Evacuating a large scale building: A Holy Mosque in Makkah as a case study

Jung-Yup Kim, Korea Institute of Construction Technology, Goyang SOUTH KOREA Development of smoke control system ensuring safe evacuation through stairwell in high-rise building in Korea

Sergey Kitov, KONE Corporation, Espoo FINLAND Improving flexibility of agent‘s path selection in cellular pedestrian flow model

Jaisung Choi, University of Seoul, Seoul SOUTH KOREA Revision of pedestrian levels of service values based on pedestrian conflicts

Huan-Huan Tian, Shanghai University, Shanghai CHINA The lattice hydrodynamic model considering the human subconscious behavior for the counter flow

Chan-Sol Ahn, Korea Institute of Construction Technology, Goyang SOUTH KOREA A study for ventilation capacity of large enclosure considering real fire load

New wayfinding techniques in pathfinder and supporting research

Charles Thornton, Thunderhead Engineering, Manhattan KS USARichard O‘Konski, Thunderhead Engineering, Manhattan KS USABryan Klein, Thunderhead Engineering, Manhattan KS USABrian Hardeman, Thunderhead Engineering, Manhattan KS USA

Inthispaper,wewilldiscussimprovementstoPathfinder‘swayfindinganddoorselectionalgorithms.Wewillalsodiscussthecharacteristicsofresearchworkthatmakeitpossibletovalidatesuchimprovements in simulation software.

Pathfinderisacommercialagent-basedemergencyegresssimulator.Thesoftwareincludesauserinterface,simulator,and3Dvisualizationsystem.Thewayfindingdiscussedinthispaperisanaspectofthesimulator.Occupantsmovingfromtheirstartinglocationtoanexitmustchoosearoutetofollowwhenwalkingtowardtheirchosenexit.Thisrouteselectionprocess(wayfinding)affectstheoverallsimulationresultsgreatlybecausetimespentwaitinginqueuesandtimespentwalkingcontrolthetimeittakesalloccupantstoreachtheirobjectives.

Previously,Pathfinderusedasimpleproceduretoperformthiswayfindingtask.UsingtheAStar(A*)searchalgorithm,Pathfinderwouldcalculate,foreachoccupant,theshortestpossiblepaththeoccupantcouldtaketoexitthemodel.Becausethisapproachdidnothingtoaccountforqueueformationatdoorways,Pathfinderusershadtouseavarietyofworkaroundstoencourageoccupantstowalkafewextrastepstoavoidlargequeuetimes.

ThenewwayfindingalgorithminPathfinderattemptstomoreaccuratelysimulatetheprocesspeopleusetoperformwayfindinginreallife.Thisnewversioncombinesdistanceestimateswiththecrowdingobservedintheoccupant‘scurrentroom.Theresultofthisnewapproachisthatoccupantsareabletore-routebasedonestimatedqueuetimesandselectmorerealisticroutesduringthesimulation.

Toimplementthisnewwayfindingalgorithm,wehadtoovercomeavarietyofunforeseenchallenges:backtracking,re-entrantpaths,performance,validation,andothers.Thispaperwilldiscussthesechallengesanddescribetheirimpactonthefinalwayfindingalgorithm.

Validationplayedakeyrolethroughoutthedevelopmentofthesechanges.ThispaperwillalsodiscusshowworksuchasDonaldHavener‘sobservationaldataonpedestriansastheyenteredanassemblyspaceandArminSeyfried‘sworktoquantifyingmanyelementsofhumanmovementwasvitaltoensuring that our simulator continued to reflect valid science. It will also discuss the characteristics that madetheseworksparticularlyhelpfulinthesimulatordevelopmentcontext.

An innovative evacuation system for multiplex cinemas: Case Study “Village roadshow group of companies - Athens“

Stavros Katsoulis, Village roadshow group of companies, Athens GREECEAthanassios Kosmopoulos, Audit and Quality Assurance Officer, Athens GREECE

Thispaperaddressestheorganizational,proceduralaswellasthetechnologicalissuesrelatedwithaninnovativeevacuationsystemforVillageRoadShowGroupofCompaniesinAthensGreece.Thepaperdepictstherequirementsneededforthiscasestudysystemtoensureproperandsafeevacuationroutesfor the public in a multiplex cinema.

Awelldesignedevacuationsystemshouldproduceclearopticaldirectionalsignalinginordertoguidetheaudiencetowardstheassignedexitdoorsandroutesincaseofanemergency.Statisticaldatashow,withrespecttoaudiencebehavioralmovements,that60%ofallaudienceinanemergencyprefersto follow the previous entry routes rather than the indicated exit ones. In such a case serious injury probabilityoccurswiththeconflictingroutingofenteringandexitingaudience.Thesystemrespondssignificantlytotheaforementionedrisk,ensuringtheintegrity,andsafetyofbothpeopleinvolvedandinfrastructure.Itisnottobeneglecteda5%ofallaudiencesrefusingtomoveinanemergencyprocess,and the proper system should motivate them towards the proper direction.

Thepaperdescribestheproceduresdeployedfortheoperationandsupportofthesysteminvolvingthehumanresourcefactorsintermsoftraining,monitoringandcontrol.Ascriticalfactoremergestheneedforimmediateresponseofthesystempartiallyortotallyinthemultiplexasfarasdecisionmakingprocessisconcerned.Thepaperdiscussesandanalysestheplethoraofemergencyrisksimposedinamultiplexcinema(fire,earthquake,bombthreat,hostagesituationetc)andthedifferentrelatedtypes of evacuation process should be followed respectively. It assesses and forecasts the impact of an evacuation failure both to the company continuity as well as of any collateral damages.

Thecasestudyincludesasoftwaresimulationofdifferentevacuationsituationsregardingdifferentaudiencepresenceinthemultiplex.Conclusivelythepaperfocusesonthespecificattributesanevacuationsystemformultiplexesshouldrespectinordertoensureaccuracy,verifiability,accountability,securityandqualityassuranceinrelationtotherelevantinternationalstandardsinforcesuchasISO.

Simulation models of merging priorities in staircases

Henri Hakonen, KONE Corporation, Espoo FINLANDMarja-Liisa Siikonen, KONE Corporation, Espoo FINLAND

Whenevacuatingahigh-risebuilding,twostreamsofevacueesmeetatthelandings:thosewhoarealready descending the stairs and those who are about to enter the staircase through the landing door. Inthisarticle,mergingofthesestreamsisstudiedwithagent-basedsimulationsusingbothqueuingmodel and cellular automata.

Thequeuingmodelhasafirstinfirstout(FIFO)queueoneachfloorforagentswhoareabouttoenterthestaircase.Thelandinghassufficientlyspacefortheagentstoenteraccordingtothemergingpriority.Thenumberofagentsinthelandingislimitedbyamaximumdensityandtheflowislimitedbyamaximumflow,whichdependsonthewidthofstaircase.Inaddition,thelandingdoorwidthoneach floor limits the flow of entering agents.

Thecellularautomatamodelsmovementsoftheagentsindetail.Thismodeldoesnotuseanyfixedmerging priorities but the joining is a result of simulated interactions between the agents and the building.

Anevacuationofahigh-risebuildingissimulatedwithboththequeuingmodelandthecellularautomatamodel.Forthequeuingmodel,thefollowingmergingprioritiesareused:1)Streaminthestaircasehasapriority:agentswaitingatthefloorshavetowaituntilthestaircaseisempty;2)Streamfromthefloorshaspriority:agentsinthestaircasehavetowaituntilthefloorisempty;3)Bothstreamshaveequalpriority.

Themergingprioritydoesnothavemucheffecttototalbuildingegresstime.Inthefirstcase,theupperfloorsareevacuatedmorequicklythanintheothertwocases.Themergingpriorityaffectstothequeuingandwaitingtimesonthefloors.Resultsofthecellularautomatamodelarecomparedtoqueuingmodelwithdifferentmergingpriorities.Similaragentcharacteristicsareused,butinthecellularautomatamodel,theshapesandsizesoflandingsarevaried.Theeffectofthelandingshapetothe merging priority is compared.

Evacuating a large scale building: A Holy Mosque in Makkah as a case study

Wael Alhalabi, Umm AlQura University, Makkah SAUDI ARABIA

TheHolyMosquethatlocatedinMakkah,SaudiArabia,thatbeenexpandedoverthetime.Todayitisabout366,168m2andcouldaccommodatemorethanhalfamillionprayers.It’samultilevelbuilding.TheHolyMosquehasmorethanonehundredgateswhichbeusedinpeaktimetogiveworshippersabilitytoreachthefourfloorsofthebuildings.Severalactivitybeenheldinsidethebuilding,suchasthefiveprayersthattakeplacedailywhichneedtheworshipperstoingressandegressthebuildingandgotopracticetheirnormallifewhenfinished.Inadditiontothat,Tawaf,Sai’andsomeotherritualsshouldbedoneandtakingplaceinsidethisbuilding.AnnuallyandduringtheseasonsofRamadanandHajj,theHolyMosquereacheditsoptimumcapacitywhenworshipperspresentfromallovertheworldtopracticetheirritualsduties.Theneededmovementofthoseworshippersandthedifferencesinbackground,ages,languages,…etc.needspecialtreatmentsandplans.

Inhistory,ithadbeenrecordedthattheHolyMosqueneedtobeevacuatedonlyincaseofwar,however,wehavetoprepareanevacuationplanforworshipperstopreventanyharmsduringstampedeandtokeepthemsafe.

Thispaperwilldiscusstheissueofevacuatingthebuildingincaseofemergency,andwheneverit’sneeded and to provide assembly area for worshippers. It will go throw the history and bring to the readerattentionallaccidentsthatbeenoccurandrecordintheHolyMosque.Inaddition,itwillapplysomeoftheinternationalCODtotestitsabilitytobeappliedonsuchabuilding.Inaddition,itmayprovide some alternatives to provide a safe building that could accommodate worshippers and to practice their ritual in a safe and comfortable environment.

Development of smoke control system ensuring safe evacuation through stairwell in high-rise building in Korea

Jung-Yup Kim, Korea Institute of Construction Technology, Goyang SOUTH KOREAHyun-Joon Shin, Korea Institute of Construction Technology, Goyang SOUTH KOREA

Asbuildingbecomeslarger,tallerandmorecomplexduetoindustrializationandurbanization,ittendstobevulnerabletofireandestablishmentofeffectivemeasuresforfiresafetyisdemanded.Especiallythefactthatthesmokehindersevacuationandfire-fightingactivitiesaswellasbecomesthemajorcauseoflifecasualtyemphasizestheimportanceofsmokecontrolsystem.

InKorea,asoneofthefiresafetystandardsdesignedtosecuretheevacuationofpeopleinhigh-risebuilding,NFSC501A(DesignGuideforSmokeControlSystemofSpecialEvacuationStairwellandLobby)hasbeenproposed,preventingsmokefrompenetratingintothesmoke-freeescaperoutebyraisingthepressureofthesmokecontrolzonehigherthanfirearea.

Tothatend,thepressuredifferentialsystemwhichraisesthepressureofthelobbybetweenaccommodationandstairbyusingcentrifugalfan,ductandairsupplydamperiscommonlyemployedinhigh-risebuildinginKorea.Thepressuredifferentialsystemforsmokecontrolshouldbedesignedtosustainthepressuredifferencesbetweenstair-lobbyandlobby-accommodationwithinacertainlevelsoastoenabletheevacuatingpeopletoeasilyopenthedoorsandinadditiontopreventthesmokefrompenetratingintothesmoke-freeescaperoute.

Inthisstudy,Thefieldexperimentsonpressurefieldintwobuildingsof21storiesand31storiesinsummerandwinterseasonwithregardtoon/offconditionofsmokecontrolsystemandopen/closeconditionofdoorarecarriedouttoevaluatetheperformanceofsmokecontrolsystemandevacuationlogistics.AndNewsystemofsmokecontrolispresentedtoimprovethecurrentsystemthroughtheanalysisofexperimentalresultsandtherealizationofnewideas.Theprototypeofnewsmokecontrolsystemismadeupandappliedtotestbedforanalysisofthesystem‘sperformance.

Improving flexibility of agent‘s path selection in cellular pedestrian flow model

Sergey Kitov, KONE Corporation, Espoo FINLANDJuha-Matti Kuusinen, KONE Corporation, Espoo FINLANDJanne Sorsa, KONE Corporation, Espoo FINLAND

Thecontinuoussocialforcemodelandcellularautomatamodelarethetwomainapproachesusedinsimulationofpedestrianflow.Incellularautomatamodelsastaticfield,whichstoresthedistancetoadestinationforeachcell,isnormallyusedforagent’sorientationinspace.Thismeansthatagent’snextmove is to the neighboring cell that is closest to the destination.

However,thisapproachmayresultinunrealisticbehavior.Agentstendtoselecttheshortestpathandignore alternative paths even if the shortest path is crowded and an alternative path would be better withrespectto,e.g.,timetodestination.Thereasonisthatagents’decisionsarebasedonlyontheinformationaboutthestaticfieldinthenearestcells,andinformationaboutalternativepathsissimplyunavailableforthem.Hence,theyactlikealternativepathsdonotexist.

Thisworksuggestsaneworientationmodeltomakeagentbehaviormorerealistic.Theideaistoselectthenextneighboringcelltomovetobyusingalinearcombinationofthestaticfieldvalueandavalueofdirection.Thevalueofdirectionforagivenneighborhoodcellisdeterminedbythestaticfieldvalueatalook-aheaddistanceinthedirectiondefinedbytheneighborhoodcell.

Agentbehaviorisstudiedusingdifferentpredefinedlook-aheaddistances,andbyassigningdifferentweightstothestaticfieldvalueandthevalueofdirection.Themodeliscalibratedusingdatafromanexperimentalstudy.Withtheneworientationmodel,theagentsbehaveinamorenaturalway:theshortest path is still preferred if it is free but alternative paths start to attract the agents if the shortest path gets crowded.

Revision of pedestrian levels of service values based on pedestrian conflicts

Jaisung Choi, University of Seoul, Seoul SOUTH KOREASangyoup Kim, University of Seoul, Seoul SOUTH KOREASeongmin Kim, University of Seoul, Seoul SOUTH KOREATaeho Kim, Ministry of Land Transport and Maritime Affairs, Seoul SOUTH KOREA

Thelevelsofservice(LOS)forpedestrianflowsaredeterminedwithKoreanHighwayCapacityManualinSouthKorea.AlthoughthepedestrianLOSmethodologyinKHCMhasanadvantageofhavingasimplecomputationalprocedure,thismethodologyseemstoinvolvesomeshortcomings.Forexample,pedestrian density was adopted to explain the discomfort in pedestrian flows but appears to be too insensitive to changes in pedestrian volume.

Theoretically,thevolumeincreaserequiredtodegradeLOSDtoEshouldbemuchlessthantheoneforLOSAtoB,becausethetotalnumberofpedestrianconflictsincreasesrapidlyathighvolumelevels.However,withpedestriandensitybeinginsensitivetochangesinpedestrianvolume,theKHCMmethodologyisunabletoaccuratelyexplainthepedestriandiscomfortateachLOSlevel.Infact,afieldpanelsurveycarriedoutinthisresearchrevealsthattheKHCMmethodologywillresultsinLOSvaluesthatarequitedifferentfromtheLOSvaluesperceivedbypedestrians.

Thisresearchattemptstoresolvethisproblembyrealigningthepedestriandensity-volumecurveonwhichthepedestrianLOSservicesarebased.Inthisprocess,itisfoundthatpedestrianconflictisamorereliable,sensitive,andcanexplainmoreaccuratelythepedestriandiscomfortlevelsaccordingtovolumeincrease.Subsequentlypedestrianconflictcharacteristicsareinvestigatedforvariouspedestrianflowsandtheseresultsarelinkedtopedestriandensityinordertorealignthecurrentdensity-volumecurve.Finally,arevisedsetofpedestrianLOSareproposedinthisresearch.ItisexpectedthatthisrevisionofthepedestrianLOSvalueswillcapturetheeffectsonpedestriandiscomfortofvolumeincreaseonwalkways.

The lattice hydrodynamic model considering the human subconscious behavior for the counter flow

Huan-Huan Tian, Shanghai University, Shanghai CHINAYu Xue, Shanghai University, Shanghai CHINARong-Sen Zheng, Yulin Normal University, Yulin CHINALi-Yun Dong, Shanghai University, Shanghai CHINA

Inthispaper,basedonthetwo-dimensionallatticehydrodynamicmodelconsideringthepathchangeinthebidirectionalpedestrianflow,anewlatticehydrodynamicmodelforthecounterflowisproposed.Inthismodel,thepedestrian’shabitsincludedthatthewalkersprefertowalkontheright-handsideofroad,andfasterwalkersareusedtoovertakingtheprecedingwalkersinthesamedirectionfromtheleft-handsideoftheroad,aretakenaccount.Inourmodel,pcr,pclareusedtorepresentthemagnitudeofthewalkersovertakingtheprecedingwalkersfromhisorherright-handandleft-hand,respectively.pr,plareusedtorepresentthestrengthwalkingtotheright-handsideortheleft-handside of road.

Thecomputersimulationsarecarriedout.Theupanddownboundariesareno-flux,therightandleftboundariesareperiodic.Theinitialconditionsarechosenasfollows:ρ(i,j)=ρc=ρ0=0.2att=0;thenthelocaldensitiesρ(L/2,W/2)andρ(L/2-1,W/2-1)attimet=1aresetas0.1and0.3,whereLandWisthelengthandwidthofthechannel,andtakingL=200,W=50.Duringthesimulation,wetakepr>pland pcl> pcr.

Theresultsofthesimulationsindicatethat,whenpcrclosestopcl,thespatialdistributionofthetotaldensityρtakesonregularity,andthejamappears.Whenpcrisapparentlysmallerthanpcl,thespatialdistributionofthetotaldensityρtakesonno-regularity,andthelaneformationtakeson.

A study for ventilation capacity of large enclosure considering real fire load

Chan-Sol Ahn, Korea Institute of Construction Technology, Goyang SOUTH KOREAJung-Yup Kim, Korea Institute of Construction Technology, Goyang SOUTH KOREA

Thisstudyisintendedtoevaluatethecharacteristicsofsmokespreadingandtheappropriatenessofevacuationtimeextendedbyoperationofsmokecontrolsystemduringfirewithintheundergroundspaceofthebuildingstructuredincompliancewiththesmokecontrolsystemperformancecriteriafromthelocalfiresafetystandard,whichhasbeencurrentlyappliedtothebuildingsinKorea.Usingtheheatreleaseperunitweightofthecombustibles,anumericalanalysisbothincaseofsmokecontrolsysteminoperationandthesystemnotinoperationwascarriedoutattheseveraldifferentshoppingmalls.Fromtheviewpointofsecuringtheevacuationtime,theresultswerecomparedinanattempttoassesstheappropriatenessofthefiresafetycriteria.

UvA-DARE (Digital Academic Repository) Simulation of city evacuation … · 6th International Conference on Pedestrian and Evacuation Dynamics - PED 2012 Accompanying booklet abstracts

Documents