INO 2004 - Eurocontrol

A C T I V I T Y R E P O R T

2 0 0 4

INNOVATIVE RESEARCH

E U R O C O N T R O LE X P E R I M E N T A L C E N T R E

© European Organisation for the Safety of Air Navigation

(EUROCONTROL)

This document is published by EUROCONTROL in the

interests of exchange of information. It may be copied in

whole or in part, providing that the copyright notice and

disclaimer are included. The information contained in

this document may not be modifi ed without prior written

per-mission from EUROCONTROL.

EUROCONTROL makes no warranty, either implied or

express, for the information contained in this document,

neither does it assume any legal liability or responsibility

for the accuracy, completeness or usefulness of this

information.

EURO

CO

NTR

OL

EXPE

RIM

ENTA

L C

ENTR

EIN

NO

VA

TIV

E R

ESEA

RC

H A

CTI

VIT

Y R

EPO

RT

20

04

RAP INO COUV REPRO.indd 1 13/07/2005, 14:27:26

Eurocontrol Experimental Center – Innovative Research Activity Report 2004 1

Dear Reader,

The major responsibility of the EUROCONTROL Experimental Centre (EEC)Innovative Research Area (INO) is to federate and promote innovations in allATM related fields through exploratory studies and by a lateral thinkingapproach for R&D issues with the goal to bring new scientific knowledge toATM R&D.

ACARE Strategic Research Agenda identifies main streams and potentialbreakthrough opportunities for the European Air Transport at the perspective2020+. The Innovative R&D Research Area of EEC, within the frame of theAgency’s new R&D organization, has the new responsibility:1

to investigate and coordinate the exploratory, initial feasibility studies onthe topics suggested in the ACARE Strategic Research Agenda, in additionto the EEC research reorientation;

and to provide input to the applied research and development activitiesthroughout the Agency EATM Directorates, as well as to the EATMstrategy and concept development entities. 2

Besides the two major roles depicted above, yet another role might be themanagement of an Academic Network of Excellence, supporting innovationsin ATM fields. 3

Innovations, however, are not exclusive to Innovative R&D as an entity, butare the responsibilities of all work programmes. In such perspective, a clearstrategy shall be defined to constrain the scope of the work programmes,both in the traditional investigation role as well as in the new coordinationrole for INO. This report shows a clear separation between research themes:

For own investigation role, the selected themes for the next five yearsinclude:

0. shift of control paradigm from non-synchronous to synchronousATM;

1. advanced 3D visualisation and interaction technologies for ATC;

2. understanding the nature of uncertainties in ATM and its non-empirical optimization.

For coordination role, the selected focus for CARE mechanism istechnology innovations for modern airports, for traffic management andfor control functions.

1 Advisory Council for Aeronautics Research in Europe (ACARE).2 European ATM Performance Enhancement Activities (EATM).3 Air Traffic Management (ATM).

2 Eurocontrol Experimental Center – Innovative Research Activity Report 2004

The cohabitation of management, investigation and coordination roles for thesearch for innovation breakthroughs requires obviously, and more than ever,a transparent and well-structured process for the selection, monitoring, anddissemination of all research activities. This continuous improvement activity ishowever not included herein.

I hope you do enjoy reading this report.

Pierre Andribet

Acting Director EEC


Dear Reader,

This Innovative Research Activity Report 2004 has been published shortlyafter the publication of the second Strategic Research Agenda (SRA) of theAdvisory Council for Aeronautics Research in Europe (ACARE).

In its Executive Summary this SRA states :

(p.5) “ACARE has prepared this 2nd edition of the Agenda. It is built upon thetechnical foundations of air transport that relies for progress upon the applicationof science and technology. Without research and the creation of new ways ofachieving ambitions there will be no progress. New problems, such as thedetermination not to allow world terrorism to halt or hinder development of travel,require new solutions and these in turn require new research and the developmentof systems. The growing concern for the environment points to the need for newresearch to understand the mechanisms that govern our complex globalenvironment better than we do and then to develop solutions. The continuedgrowth of globalised industrial trading requires that freight and passenger patternsare changing and new services are needed. Congestion of our fixed airport and airtraffic management infrastructure is causing massive frustration to operations andpassengers alike and needs new technologies and different ways of co-operatingto overcome.”

This clear request for “looking beyond the present horizons” is further spelledout in Volume 1 of the SRA:

(p.64) “Looking at the technologies needed to deliver the air transport system of2020 has its own challenges but is also constrained. Many of the technologies arewith us already, others will need to be developed in the short term if they are tobe delivering benefits by 2020. On the whole, therefore, the technologies discussedin the other High Level Target Concept’s are mainly evolutionary developments.But ACARE wants also to look further ahead, to the middle of the century andbeyond. It wants to explore and pioneer the more radical, revolutionary andinnovative combinations of technology that might have the power to cause aradical step change in air transport in answer to the long term challenges anddesign imperatives.”

This vision at European level has found its counterpart in the U.S.A. wherethe Joint Planning & Development Office (established by several FederalGovernment Departments, the FAA and NASA) published its “NextGeneration Air Transport System – Integrated Plan”. Like the SRA it focuseson the needed changes that can only brought along by innovation.


Jointly Europe and the U.S.A. support the development of global concepts inan ICAO setting, thus assuring the airspace users that we are converging to atrue world system.

In this global context there is full recognition that Innovative Research shouldget more attention and leverage. So the future looks bright. Let us thereforecontinue on the path we have taken and not forget that next to the freedomin exploring new frontiers, we also have the responsibility to anchor the mostpromising opportunities into a deployable strategy. And that we have to doso in close connection with all sectors of the aeronautical community andwith all stakeholders worldwide.

Jan van Doorn

Chairman of the InnovativeResearch Advisory Board


INNOVATIVE RESEARCH

INO

The arrival of two deputy research areas managers and a new managementassistant in 2004 have brought several improvements to INO. The mostrelevant achievements of the year were:

The Maastricht ATC 2004 Innovation Award, granted to the work onAircraft Identification Tag, from an original of Horst Hering, using digitalwatermarking technology to insert call-sign tags in pilot-to-ground VHFcommunications.

Third Annual Innovative Research Workshop stepped into a newdimension with notable larger audience, counting to 150 participantsmainly representatives of stakeholders from 24 member states; On thisresult dissemination activity, the First International Conference onResearch in Air Transportation, jointly organised with the University ofZilina, Slovakia also attracted 139 young scientists and researchers of 30different nationalities from 20 countries, and had been a success.

The first EUROCONTROL Joint Research Lab was concluded with theSorbonne University, Ecole Pratique des Hautes Etudes (EPHE), aprestigious research-orientated institution of France, to establish theComplex System Modelling and Cognition Laboratory at the EEC. As astep toward the network of academia-EUROCONTROL Joint ResearchLabs, the work programme of this laboratory consists in the modelling ofthe ATM system from a cognitive and complex system approach,identified in ACARE/SRA as a crucial backbone thinking topic.

The management of the Collaborative Action for Research inEUROCONTROL, (CARE) relating to Innovative Research was effectivelytransferred to the research area. The second call for proposals hadreceived 50 proposals from universities, industries and researchinstitutions, among which 6 were selected to be supported.

The four major research axes redefined in 2003 have been now confirmed asan integral part of the Centre’s work package relating to Innovative Research.

On the research axis relating to advanced operational concept, the ParadigmShift project (SHIFT) achieved its 2004 objective to provide a thorough viewof a concept of operations [INO-04-01 INO-04-03] that capitalised onprevious investigations: Sector-Less, Super-Sector, Trunk Route Networks andcity-pair highways - TALC. Based on a dual-route network, combining long-haul highways route and standard route with decentralised control, SHIFTproposed a concept of operations using the notion of time-flexible contracts


and management of interruption to deal with operational uncertainty. Thissuggestion for a shift of control paradigm for the 2020+ horizon has receivedvery positive feedback from the audience of the Innovative ResearchWorkshop. Progress has also been made on the investigation on using speedcontrol in conflict detection and resolution, new parameters relating toconflict density is suggested from this work [INO-04-05 to INO-04-07].

On the research axis relating to Advanced Technologies, remarkable progresshas been made on the investigation of 3D stereographic display for ATC: Theintegration of voice interaction into the display has demonstrated to betractable [INO-04-08], and the human-centred design approach has alsoshowed higher performances from 3D stereoscopic display with respect to2D interfaces [INO-04-39, INO-04-09]. New ideas for currently usedcontroller interfaces are also suggested to better assist controllers in theirtasks [INO-R-10 to INO-04-12].

On the research axis tackling the issues of uncertainty modelling, progress hasbeen made on the uncovering of differences between executed flights andCFMU scheduled flights. Characteristics of differences between executed andscheduled flights have been demonstrated to be time-scale significant: over-deliveries are systematic when traffic demand decreases in time scales rangingfrom 5 minutes to 15 minutes. However regularities in patterns characterisinguncertainty are still yet to uncover [INO-04-13 to INO-04-13 to INO-04-16].

Contributions to public scientific knowledge from INO amounted to 37reports and papers in international conferences, including three PhDdissertations of the first three INO PhD students [INO-04-35 to INO-04-37].Six new PhD students have been recruited in 2004, and three additional PhDgraduations are expected in the first quarter of 2005.

`

Vu Duong

Head of Innovative Research


INNOVATIVE RESEARCH ADVISORY BOARD

IRAB

Members are:

Jan VAN DOORN – Chairman

EUROCONTROL headquarter

Bernard MIAILLIER

EUROCONTROL headquarter

Jean Luc MARCHAND

European Commission

Patrick BELLOT

Ecole nationale supérieure des télécommunications

Dominique COLIN DE VERDIERE

CENA

Kevin CORKER

San José State University


INNOVATIVE RESEARCH AREA

Management Team:

Vu DUONG Research Area ManagerMarc BROCHARD Deputy ManagerMarc BOURGOIS Deputy Manager

Members of the Research & Administrative Staff:

Rüdiger EHRMANNTRAUTJean Pierre FLORENTSandrine GUIBERTLaurent GUICHARDJean Luc HARDYHorst HERINGMartina JURGENSChristian MUSSONHughes ROBIN

PhD Students : Contractors :

Magnus AXHOLT Khaled BELHACENENabil BELOUARDY Didier DOHYPeter CHOROBA Jean Yves GRAUAntonia COKASOVA Jean NOBELThong Nguyen DANGFrédéric FERCHAUDClaus Peter GWIGGNERKonrad HOFBAUERHa Hong LEMartin MATASSteven PETERSONElzbieta PINSKADaniel ROHACSMonica TAVANTIThomas RIVIERE (off-site)Nicolas ARCHAMBAULT (off-site)Devan SOHIER (off-site)


TABLE OF CONTENTS

Innovative Research INO 5

Innovative Research Advisory Board IRAB 7

Innovative Research Area 8

ADVANCED CONCEPTS 13

Analyses of Passengers’ Preferences in Air-rail Intermodal Transport 21Antonia Cokasova, Vu Duong 21

The Airport of the future or breaking the constraints between the terminaland the runways 37

Marc Brochard 37

The Airport of the Future or what can be the airport in the year 2020 andafter ? 47

Martin Matas, Marc Brochard 47

Project "Paradigm SHIFT" 53Laurent Guichard, Sandrine Guibert, Horst Hering, Jean Nobel, Didier Dohy, Jean-Yves Grau,Khaled Belahcène, Marc Brochard 53

FAME Future Air Traffic Management Performance Enhancement 67Gilles Gawinovsky, Marc Brochard 67

Tube Advanced Lane Control 77Jean-Pierre Florent, Marc Brochard 77

Full Automation of ATM for High-Complexity Airspace 89Rüdiger Ehrmanntraut 89

Analysis of the Impact of Small Aircraft on ATM in Europe 95Daniel Rohacs, Marc Brochard 95

EMERGING TECHNOLOGIES 109

Digital Signatures for Air-Ground Voice Communications 115Konrad Hofbauer, Horst Hering 115

3D VR Air Traffic Management Project 121Marcus Lange, Matthew Cooper, Anders Ynnerman, Vu Duong 121

Introduction and Evaluation of Selection-by-Volume Approach 133Nguyen-Thong Dang 133

Augmented Reality Tools for Tower Control 153Magnus Axholt 153


Visual Augmentation of Airport Objects in Air Traffic Control Towers 157Stephen Peterson 157

Air Traffic Tower Controller’s Tasks Performance in Augmented Realityenvironment 163

Elzbieta Pinska 163

Vital – Advanced Time-Line Approach for Future ATM Environments 169Horst Hering 169

Wheelie – Mobile Horizontal Display Filter to Ease Controller’s SeparationTask. 177

Horst Hering 177

ATM MODELLING 183

Implicit Relations in Flight Data 185Claus Peter Gwiggner 185

Optimal Path Planning for Air Traffic Flow Management under StochasticWeather and Capacity Constraints 189

Alexandre d’Aspremont, Devan Sohier, Arnab Nilim, Laurent El Ghaoui, Vu Duong 189

Towards a Next Generation ATM System Model Based Conflict Detectionand Resolution 201

Dr John Lygeros 201

Absorption Areas Utility in ATFM 207Frédéric Ferchaud, Cyril Gavoille, Mohammed Mosbah, Vu Duong 207

Performance of Air Traffic Flow Management 219Nabil Belouardy 219

Quantitavely Estimating Wake Vortex Safety using P2P Model 229Yue Xie, John Shortle, Peter Choroba 229

Column Generation for Dynamic Short Term ATFM 249Olivier Richard, Rémy Fondacci, Wojciech Bienia, Maurice Queyranne 249

Optimal Flight Level Assignment: Introducing Uncertainty 261Sophie Constans, Nour-Eddin El Faouzi, Olivier Goldschmidt, Rémy Fondacci 261


INNOVATIVE RESEARCH WORKSHOP 277

Eurocontrol Innovative Research Workshop 278

First Eurocontrol Joint Research Lab 281

Eurocontrol Innovative Research Workshop Exhibition 283

COOPERATIVE ACTIONS OF R&D IN EUROCONTROL 285

Enhancement of AGT Telecommunication Security using QuantumCryptography 289

Ecole nationale supérieure des telecommunications 289GET & LTCI-UMR 5141, Paris, France. 289

VisuAirport Adapted Observation for Activities of an Airport 303ARMINES, Ecole des Mines, Paris, France. 303Readymade, Paris, France. 303

SCOPE Safety of Controller-Pilot Dialogue 309Thales Research & Technology France. 309Thales Corporate R&D centre, Orsay, France. 309IntuiLab, man machine interfaces SME,Toulouse, France. 309IRIT, Institut de Recherche en Informatique, Toulouse, France. 309

ANIMS Improving the efficiency and safety of ATM user interfaces with visualanimation and sound 315

IntuiLab, Man Machine Interfaces SME,Toulouse, France. 315Intactile Design, Graphic design and Sound design SME, Montpellier, France. 315

Neural network-based recognition and diagnosis of Safety-Critical Events 325National Aerospace Laboratory NLR, The Netherlands 325Foundation for Neural Networks, University of Nijmegen, The Netherlands 325

The Airport of the Future Central Link of Inter-Modal Transport ? 331M3 SYSTEMS, Space and Aeronautics SME, France. 331ENAC, Aviation Economics and Econometrics Laboratory, Toulouse, France. 331ANA, Aeroportos de Portugal S A, Portugal. 331

CREA Training 337DeepBlue, research and consultancy SME, Italy. 337

INO PUBLICATIONS 343


AdvancedConcepts

Advanced Concepts


The Advanced Concept threadaims at envisioning and evaluatingadvanced concepts for drivingcontinual growth seen as the majorconstraint to the future AirTransport System. The mainconclusion of the work undertakenin the different fields is that theVision 2020 objectives would notbe reached unless there was aparadigm shift in the way the AirTransport System is conceived andoperated. Following the ACAREStrategic Research Agenda (SRA 2)recommendations, this threadrecognises the need of envisioningnew paradigm in the field of airportof the future, of advanced AirTraffic Management system, ofhighly automated ATM and ofimpact of new air vehicule on ATM.All these activities must be steeredby the need of sustainable airtransport business development.During the year 2004, all activitiespart of this thread have been fullyre-aligned accordingly.

The Airport of the Future

Considering the foreseen trafficgrowth (doubling or tripling oftraffic in the 15 coming years), therequired f l i ght punctua l i tyimprovement, and the neededreduction of time spent in theterminal area by the passengers, asstressed in the ACARE report;airport has been identified as one ofthe major element which willconstraint this traffic growth if nomajor improvement is undertaken.

Researches are ongoing in order tobetter understand the airportproblematic (especially passengerexpectations, including inter-modality issues), and to envisionwhat could the airport be in thefuture seeking to break thecontsraints between the terminaland the runway, and the result intonew concept of networks ofairports integrated in an overall air-rail multi-modal transport system.

------- High Speed Train ---------

Future airport paradigm Two PhD activities have been

performed in that field. The firstone is the continuation of researchinvestigating passenger travelpreferences from a perspective ofair rail competition in order toanalyse passenger modal split inEurope in the future and todemonstrate the necessity andimportance of High-Speed Traininter-modal connections betweenEuropean airports. This work hashighly participated to the lead to anextreme knowledge improvementThe second one has started theexploration of new airport

Advanced Concepts


paradigm based on breaking theconstraints between the terminal(the airport land side) and therunways (the airport air side). This isa challenging work which tacklesmany issues such as, for example,airport business, runway operation,transport network (ATM and inter-modality), airport infrastructure andintegration with the overall ATM,inst i tut ion , economics andenvironment. This innovativeconcept might not be translatedinto operation in the future, but wedo expect it will enrich ourknowledge in the airport domain,which could lead us to proposeother alternatives for improvingcapacity at the airport level. Theseactivities have been linked withsome CARE INO activity which, alltogether, could draw inter-modalitypicture in the future and shape theairport of the future.

Advanced ATM concept

As stressed by ACARE report andmany other bodies, current ATMsystem will not be able to copewith traffic increase planned overthe years to come, neither tosupport a sustainable air transportbus ines s . Researches a reundertaken in order to envision theoperational concept for ATM of thefuture embracing airspace, system,automation and human in the loop,this keeping in mind safety as beingthe first objective over any other,and the need to increase capacityand efficiency. The philosophy ofthis operational concept is that

different modes of operation willprevail in different parts ofEuropean airspace. Modes willcorrespond to different “qualities ofservice” related to airspace userneeds, levels of aircrew, aircraftembedded equipments and trafficdensity. These researches focus onthe development of requirementsfor such a system, operationalconcepts, safety and cost-benefitaspects and demonstrations.

In that context and keeping in mindthe underlying philosophy, theParadigm SHIFT project aims atdeveloping an operational conceptto optimize the use of resources byreducing the uncertainty and aimingat offering a new way of managingthe ATM process (with all actorsdealing with ATM such as Airport,Airlines, and ANSP). Consideringthe Air Navigation Services as a partof the Air Transportation System,global approach of this compositesystem has been adopted. Safetyand efficiency issues are addressed.Year 2004 has been one yearthinking about developing a newand innovative advanced ATMconcept. A group of several expertscoming from ATM domain,operational air traffic controllers(ATCo), human factor and systemarchitecture went through severalbrainstorming sessions and finallycame up with five main interrelatedoperational concepts: Contract ofObjectives, Operational Plan, TargetWindows, Decentralized Airspace,and Dual Airspace. The analysisintroduces a new way of designing

Advanced Concepts


the air navigation infrastructurebased on the concept ofmanagement by objective instead ofby means. It defines the foundationof an ATM system able to copewith traffic demand at the horizonof 2020 and beyond, whilemaintaining a very high target levelof safety and supporting asustainable Air Transport businessdevelopment.

These concepts required furthertheoretical , operational andsustainabi l i ty analysis . Somequestions like responsibility issues ofthis highway implementation, ormore quantitative assessment ofsome parameters like trafficthroughput, should be added. Thequestions addressed by these fivemain concepts have been listed inthe form of a Research Agenda.Investigation topics are identifiedand structured into interrelatedstand-alone experimental researchtopics, with a multidisciplinaryapproach. This new concept is quitepromising and several tasks havebeen launched to validate it. We dobelieve that this paradigm shiftagenda is becoming the backboneof the advanced concept thread,meaning that any work on advanced

concept will be linked to theparadigm shift research agenda.

Early the year, an other advancedconcept has been proposed to EC6FP Aeronautics and space secondcall: Future Air Traffic ManagementPerformance Enhancement (FAME).The strategic objectives addressedthrough this proposal were topropose an innovative ATMsolution able to face the challengeof traffic demand (2020 horizon),and to improve the efficiency andsafety level of the European AirTransport System as stated in theACARE SRA. This was proposed tobe achieved through the adoptionof a time based system architectureand p roces se s p rov id i ngsynchronisation between the ATMchain components. This conceptwould lead to a more efficientnetwork and flow managementbetween high-density airports. Thekey features to exploit this conceptare a dynamic layered planningproviding the basis of 4D contractsfor aircraft trajectories, a newconcept of automated supporttools and system operator tasksharing for Air Traffic ControlWorking Positions. This 4Dcontract will thus be negotiated and‘signed’ between aircraft operatorsor the aircraft itself, and the AirTraffic Management system and anyother relevant actors. In thesecontracts, 4D trajectories will beassigned to the aircraft to ensure anacceptable level of traffic complexityand density. Changes to thecontract will be managed either

Advanced Concepts


automatically by the issuance of anew contract, or by specific actionsthat will ensure minimal impact onthe traffic situation. This concepttakes us from an essentially non-synchronous and conflict-basedsystem constraint by the tacticalcontroller cognitive capabilities to abetter synchronized (using allavailable capacity in a more effectivemanner) and highly automated(providing tighter, more accuratecontrol loops) system providingflexible and adaptive mechanisms toadjust the predictive part. Even ifthis proposal has not been selected,several concept which wereproposed to be further developed,are still under consideration withinthe INO department.

The Highly Automated Air TransportSystem

As stressed by ACARE andITA/ERASMUS reports, significantsystem performance improvementscan be achieved with higherautomation levels, i.e. an increase ofcapacity beyond levels of humanoperability with guaranteed andpredictable safety levels by reducingthe costs of the overall ATMsystem. Research is undertaken inorder to envision highly automatedair transport system. The researchfocuses on the development ofrequirements for such a system, itsoperational concept, safety andcos t -bene f i t a spec t s anddemonstrations. In that context,several tasks are ongoing.

The Tube Advanced Lane Control(TALC) concept is proposing a newconcept of highway aiming atimproving the efficiency of the AirTransport maintaining the safety in adense traffic area. This conceptdevelops a new airspace design anda new operational mode in areserved area: the tube. A newdistribution of tasks and roles isproposed between the main actorsof the air transport: i.e. Airport,Airline, Pilot, ATCo and theirrelations with automated systems.The project assesses also the impactof a collaborative decision makingapplied to the air fully derived fromthe CDM airport concept: newdata, new processes for betteradapting resources to the trafficdemand.

In the same field of developing thisconcept of highways, INO is alsoinvolved in project submitted to theEC 6FP aeronautics and space, thirdcall by the end of 2004. Thisproposal is based on the innovationof the operational framework ratherthat on the development of newtechnologies. Emphasis is placed onobtaining the maximum synergiesbetween locally available systemsand the overal l Europeanframework, thus optimising theoverall efficiency of the system.Special attention will be placed onguaranteeing the accessibility andequity of the proposed airspacestructure. The project proposes todefine and validate (using acombination of fast-time andhuman-in-the-loop simulations) a

Advanced Concepts


set of Operational ConceptScenarios that fully use EurocontrolOCD principles to arrange themajor traffic flows and patternsusing pre-defined routes, to ensurethat aircraft fly more safely andmore frequently along a “SuperHighway”. The traffic structure willbe within the Single European Skyfunctional blocks of airspace and willhave two essential components:entries/exits onto the superhighway (junctions) and lanes. Thiswork is proposed with severalpartners: ISDEFE, AENA, DFS,EUROCONTROL ExperimentalCentre (INO department) andSENASA.

The Full automation of ATMconcept is adressing studies toincrease sector capacity in highlycomplex airspace by improving themedium term planning function bybetter traffic organisation and byde-complexification of traffic withthe help of algorithms. Investigationsof traffic patterns, conflicts andcomplexities are undertaken as wellas studies on potential degrees offreedom, e.g. speed control, offsetand so on.

With similar objective of developingautomation in the ATM, theERASMUS project has beenproposed to the EC 6FPaeronautics and space, third call bythe end of 2004. The strategicobjectives addressed through theERASMUS project are to proposean innovative ATM system able toface the challenge of traffic demand(2011+ horizon), and to improvethe efficiency and safety level of theEuropean Air Transport System asstated in the ACARE StrategicResearch Agenda. At the strategiclevel, the project proposes todevelop subliminal application,when at the tactical level, ATCautopilot and enhanced MTCDapplications would be proposed.This work is proposed with severalpartners who are Eurocontrol,DNA/CENA, Honeywell, Universityof Linköping, University of Patrasand SICTA.

ATM for new vehicle

Considering the emergence of newvehicles, such as UAV’s (UnmannedAir Vehicle) or “individual”airplanes, ATM system will beimpacted even if new airspacemanagement paradigm mightprevent of having too muchheterogeneous types of aircraft to

Advanced Concepts


be controller in the same airspace.Researches are undertaken in orderto envision specific airspace andcontrol procedures for UAV’s, aswell as simplified if not automaticcontrol for individual aircraft flying inthese areas. The Small aircraftimpact on ATM in Europe study isaddressing the impacts as well asthe potential solutions regarding thegrowth of small business jets andUAV in European ATM system.

One PhD activity has just startedaiming at analysing the impact ofsmall aircraft on the ATM inEurope. Starting by a depth analysisof the current situation and theoverall foreseen evolution, this taskwill propose evolution of thecurrent ATM system, or definition

of a new one able to cope with thiskind of new aircraft.

Even if these threads areinvestigating different conceptsranging from a complete new ATMsystem to new concept applicablelocally only with strong investigationinto automation, working on bothland, ground and air sides, all theseactivities are strongly inter-relatedand close coordination is performedin order to steer toward a commonobjective: improving the AirTransport system efficiency andcapacity assuming a huge increase interms of traffic demand and trafficdiversity in the overall context ofsustainable Air transport businessdevelopment.

Analyses of Passengers’Preferences in Air-Rail Intermodal Transport


ANALYSES OF PASSENGERS’ PREFERENCES IN AIR-RAIL

INTERMODAL TRANSPORT

Antonia Cokasova(1,2), Vu Duong(1)

(1) EUROCONTROL Experimental Center, Brétigny, France.

(2) University of Zilina, Slovakia.

The objective of our research is to investigate passenger travel preferencesfrom a perspective of air rail competition in order to analyse passenger modalsplit in Europe in the future and demonstrate the necessity and importance ofHigh-Speed Train inter-modal connections between European airports.

There are numerous advantages in transferring some short haul flights to high-speed train principally that it releases runway and ATC resources, offersimmediate relief to congestion, reduces negative environmental impacts, andfinally improves ground access to airports. Passenger perspective is the keyelement; it is rather impossible to develop a well-organised and satisfactoryinter-modal interchange node with efficient baggage handling logistics andintegrated ticketing if there is no passenger feedback. In order to provide thelatest results of our research this paper will focus on detailed analyses ofpassenger travel preferences and some existing barriers of inter-modality frompassenger perspective.

Our approach was to investigate passengers open to air-rail competition inorder to better understand passengers’ perception of the potential for modechange and their main requirements. For this reason we focused onpassengers travelling on board of Eurostar (London-Paris), Thalys International(Paris-Amsterdam) and passengers travelling on short-haul flights at LisbonInternational and Paris Charles de Gaulle Airport. All passengers have agreedthat ticket price, travel time and access to the airport or station are the mostrelevant travel attributes based on which they decide on their mode of travel.

1. Introduction

Co-operation between air and rail transport, or inter-modality in other words,is a combined air rail journey that results into a seamless travel experience.Despite the fact that some barriers of inter-modality are highly visible thereare number of reasons to support the idea of system integration. Some of themain reasons are the high potential growth of air travel, congestion in air andon the ground, delays and their rising cost to operators. In Europe 10% of the

Advanced Concepts


city pairs represent as much as 50% of the traffic, due to airlines’ hub andspoke way of operation. En-route ATFM delays have significantly decreased insummer 2003 (1.2 min/flight, -35%), met the agreed target (2.1 min/flight),and nearly reached the medium term optimum delay target (1 min/flight).However airport ATFM delays increased in proportion and have notimproved significantly in absolute value since 1997. The distribution of ATFMdelay duration continued to improve for en-route but not for airports. Since2000, airport ATFM delays have doubled in proportion (46% vs. 23% of allATFM delays). In total 8 airports (Frankfurt FRA, Rome FCO, Paris CDG,London LHR, Milan MXP, Zurich ZRH, Amsterdam AMS, Munich MUC,Barcelona BCN, Vienna VIE, Heraklion HER) account for 19% of arrivals inEurope, generate 71% of all arrival airport ATFM delays and 31% of the totalATFM delay [1].

2. Understanding travel preference rules

Individuals choose to travel by a mode of transport that offers a preferredbundle of levels of attributes which are important in making the choicebetween available alternative transport modes. In determining travelpreference rules, individuals implicitly attach weights to a set of attributes thatinfluence their choice, and make a choice based on the available set. Thechallenge is to identify these weights and in so doing obtain knowledge ofwhat attributes drive an individual’s choice. An attribute with a very lowweight would be unimportant. To complete the set of items needed toderive a demand function a questionnaire was designed to identify thehomogeneity of passengers; main passenger groups and major travelattributes that most passengers find crucial when deciding between air and railtransport.

3. Objective

One of the objectives of our research is to gain a sound knowledge ofpassenger requirements; a weight to determine the most important travelattributes related to inter-modal transport and assign an importance to eachattribute. A general unknown in the field of inter-modality is passengerbehaviour.

Lessons learned from the passengers will help to obtain a forecast ofpassengers’ tendency to shift to different transport modes consideringinfrastructure network, in order to evaluate the potential air traffic reductionbetween certain city-pairs as a result of a modal shift. The main objective is toforecast the impact of inter-modal transport on air traffic, consideringpassenger requirements, high-speed train infrastructure and transportoperators’ vision.



4. Travelling by HST – up to what distance ?

In the case of air travel passengers spend at least 3 hours by travelling to theairport, waiting at the airport and checking-in. Travelling by rail the time spentreduces to 1 hour 10 min. Airports are moving further and further from thecities they serve, a reflection that airports are not good neighbours, with noiseand pollution being among the most significant problems. But the move awayfrom city center’s brings more problems – notably that of access. Differentcities have different public transport and road networks. The time needed toreach an airport can be anywhere between 40 to 120 min in extreme cases(Paris CDG). For a particular city pair, even if the time spent on a train ismuch longer than the time spent actually in the air, due to the difference inwait and access times there is a threshold distance where the total journeytime is shorter if the journey is undertaken by rail [2].

The journey duration of each transport mode indicates that high-speed trainscould replace flights of up to 750 km’s [2]. Although this distance isconsidered to be short-haul in the aviation business, in Europe the catchmentarea of 750 km’s can connect significant origin destination pairs as seen below.Out of the 20 busiest routes in Europe, 9 are above 1000 km, 3 routes arebetween 800-900 km and 8 are less then 800 km. In theory high-speed traincan replace 40% of the 20 busiest routes. For passengers that are lesscautious about time the percentage rises to 55%. Examples show that HSTcompetes with air services on routes of 300-600 km distance. Naturally theshift to rail by passengers decreases as distance grows. Most studies talk aboutdistance from 500 to 800 km [2, 3, 4, 5], very much depending on passengers’sensitivity to different travel factors. However there is more to a journey thana simple equation of time, distance and speed. There is baggage to betransferred, tickets to be exchanged; quality of interconnection points, there ispassenger comfort, safety & security and many other factors, becoming crucialdeciding factors when it comes to passengers’ choice of travel mode.Passenger shift to high-speed rail depends on the level of satisfaction, notnecessarily providing better services than air but fulfilling certain needs thatstand in priority to others.

5. Questionnaire design and methodology

A questionnaire is not some sort of official form, nor is it a set of questionswhich haven been casually jotted down without much thought. We shouldthink of the questionnaire as an important instrument of research, a tool fordata collection. A questionnaire has a job to do: its function is measurement.

But what is to measure? The answer to this question should be contained inthe questionnaire specification, which can mean many weeks of planning,

Advanced Concepts


reading, design and exploratory pilot work before any sort of specification fora questionnaire can be determined.

Before we made a start with our questionnaire some important points had tobe tackled. Are we conducting a short, factual inquire or are we conductinganalytical research on a set of attitudes? How large is the sample likely to be?Shall we be dealing with adults or with children? If with adults, will they behousewives, company directors, relatives of prisoners, students or probably ahealthy mix of the entire population? All these, and many other issues, willeffect our measurement specification and procedures.

Each survey has its own particular problems, but it is possible to presentsome general considerations that have to be borne in mind in most surveysand about which decisions will have to be made before we can begin to writeour first question. These decisions fall into five groups:

1. The main type of data collection instruments which will need, such asinterviews, postal questionnaires, content analyses of records,observational techniques and so on;

2. The method of approach to respondents (after their selection throughthe sampling procedures), including sponsorship, stated purpose of theresearch, length and duration of our questionnaire, confidentiality andanonymity;

3. The build-up of question sequences or modules within the questionnaire,and the ordering of questions and scales or other techniques within ageneral framework;

4. For each variable, the order of questions within module, usingapproaches such as personal information (age, gender) at the end of thequestionnaire;

5. The type of questions to be used; for example ‘closed’ questions withprecoded answer categories versus free-response questions.

Each of these topics had been discussed, bearing in mind that every survey isunique to a large extent. The best way to move from general to theparticular, in order to find solutions for specific dilemmas, is through small-scale field trials. Guess-work, intuition, expert advice and spurious orthodoxyare no substitutes for properly conducted pilot work [7].

Regarding type of questions and responses we have decided to use severaltechniques. Scales are used to obtain responses that will be comparable toone another, and scales can be arranged so they capture answers to manyquestions quickly and in vary little space. One of the examples of scaling weused in our questionnaire is the verbal frequency scale. This form of attitudemeasurement was ideal in order to obtain people’s position on certain issues



and conclusion. The verbal frequency scale contains five words that indicatehow often an action has been taken.

The ordinal scale is actually a simple choice item that shares some of thearithmetic characteristics of a verbal frequency scale. With a multiple-choiceitem, the response alternatives don’t stand in any fixed relationship with oneanother.

Forced ranking of items produce ordinal values, just as the verbal frequencyscales and the ordinal scales do, only the items are each ranked relatively toone another. The forced ranking scale obtains not only the most preferred,but also the sequence of the remaining items. In our case people are facedwith choices among travel attributes, they are constantly making choicesamong a limited set of options.

In other case we wanted to measure simple choice between alternatives. Thepaired comparison scale is appropriate to such a need. The focus must bealmost exclusively on the evaluation of one entity relative to another. In thissense, paired comparison can be regarded as a special case of ranking, whereonly two items are ranked at a time.

There is seldom a single, clear-cut choice of a scale for any given question orinformation requirement when composing a questionnaire. Thus, it isimportant to list a set of rules that dictate exactly what scale should be usedin each situation, even if every circumstance could be anticipated. Yet clearly,some scales are easily identified as potential tools for some commoninformation needs and questions, and there are often other scales that areclearly inappropriate.

When designing our questionnaire we tried to view the questionnaire itself, inthree main parts; the introduction, the body of the questionnaire, and theconclusion. In the first part we described the purpose of the inquiry andincluded only questions that are fairly quick and easy to answer, avoiding anyquestions that may be delicate or sensitive to respondents. The respondentsget the feeling they have done a lot very quickly and easily. Once started, theyare likely to continue. The body of the questionnaire is the middle part and itcontains questions that deal with the substance and detail of the surveytopics. The third part was reserved for kinds of questions that measure theattributes and characteristics of the respondents, demographic and biographicquestions. The reasons for putting these questions last are compelling. By thistime the respondents have become familiar with the inquiry, they have moretrust and are less likely to be sceptical or uncooperative than at an earlierpoint. Second, some respondents may terminate at this point or refuse toanswer some of the items. Nevertheless they have provided the bulk of thedata and their responses to the earlier items in the body of the questionnairemay still be usable.

Advanced Concepts


In order to address all the passengers we have created two versions, separatefor leisure and business passenger both containing 14 questions. Thequestionnaire had to be available in English and French version. Another stepforward was collecting and processing data and to decide what method willbe used to approach the respondent. The target population was clear fromthe beginning; the aim was to address passengers open to air/rail competition,given the choice to travel either by high-speed train or airplane. The only wayto succeed and collect significant number of responses was a self-administered questionnaire, when the questionnaire is presented and brieflyexplained to the respondent by someone and then the respondent is leftalone to complete the questionnaire. This method of data collection ensureda high response rate, accurate sampling and a minimum of interviewer bias,while providing necessary explanations (but not the interpretation ofquestions) and giving the benefit of a degree of personal contact.

6. Thalys & Eurostar / Lisbonne & Paris Roissy CDG

The aim of the questionnaire was to address passengers exposed to thechoice to undertake their journey by either high-speed train or airplane.

The two main railway operators were contacted during an early stage of thequestionnaire design. Both Thalys International and Eurostar have grantedauthorization to undertake the questionnaire on board of trains for the lengthof 4 days on Paris-Amsterdam/Brussels and on Paris – London route. Wehave agreed that business passengers will be approached half an hour beforethe end of the journey so they can enjoy their meal without beinginterrupted. We can conclude that both Thalys and Eurostar passengers werevery collaborative. It was easier to approach passengers travelling on earlymorning trains, afternoon passengers were more reluctant to cooperate.

Questionnaires were designed and analyzed using software called PERSEUS,dedicated for questionnaire design and analyses. It proved to be a very userfriendly, easy to learn and undertake any changes. The only big drawback ofthe software was the data input. Since it was a survey conducted via email,we were not able to fetch the responses automatically. Questionnaires had tobe handled as an interview, which required a manual input of all responses tothe database as if filling out more than 900 questionnaires; an extremely timeconsuming task.

In case of air transport the form of the questionnaire had to be adjusted,ensuring that the time to fill out the questionnaire is significantly shorter. Wehave focused on key questions identified from our experience with high-speed train passengers. Most of our attention was focused on terminal 2B and2D because the majority of the flights at these terminals are short-haul.



Airport questionnaire distribution was different from railways, due todynamic environment and additional stress in airport terminal. Questionnaireshave been collected both is boarding gates and check-in areas (for detaileddata on response rate and incomplete samples see table below).

Sample Responserate

Incompletesamples

Valid samples

Thalys 455 68% 19% 260

Eurostar 436 74% 17% 276

Paris CDG 260 87% 5% 215

Lisbon 206 85% 8% 162

7. Rail Passenger Preferences

According to prior analyses [2], we have assumed that the most importanttravel attributes that influence passenger choice of air and rail were:

ticket price

travel time

access to airport or station

schedule & frequency

punctuality & reliability on-board comfort

luggage handling

For more detailed analyses we have compared several categories ofpassengers: business with leisure passengers; frequent with not frequentpassengers; different genders and different nationalities.

Figure 1 show the statistics performed on the responses collected on Thalysand Eurostar. According to responses considering the entire samplepopulation, there are three major categories of importance that affect choiceof transport mode:

the first category of attributes that more than 60% of the populationassigned as very important were; ticket price, travel time, access to theairport or station,

the second categories of attributes with certain significance were; comforton-board, schedule & frequency and walking & waiting time,

the third category of attributes that proved to have little or no impact onpassengers’ choice between travel modes were; on-board services andluggage handling.

Advanced Concepts


Figure 1 – Comparison between Thalys and Eurostar passengers.

There is a significant difference in travel factor preferences when comparingbusiness with leisure or frequent with not frequent passengers (as Figure 2shows more that 87% of the frequent travellers were travelling for business).As seen on Figure 2 the breakdown between business and leisure passengerswas practically equal.

Figure 2 – Railway passengers.

For passengers travelling only on few occasions a year, price was a veryimportant factor, as opposed to frequent travellers that assign moreimportance to time, access to station and comfort on-board.



Figure 3 illustrates travel preferences between genders and the resultsresemble the difference between frequent and not frequent passengers.Women find price more important than men, they are ready to trade areasonably cheaper ticket for a longer journey and slightly less comfort.

It is interesting to notice that almost 10% of women are afraid to fly. Despitethe fact that it is not a high percentage, fear can be a strong incentive forchoosing rail transport. One of the most counter-intuitive factors was theluggage handling. We assumed that railway passengers are les sensitive aboutit, hence choosing rail transport, but airline passengers might consider luggagehandling as an important factor. Our assumptions proved to be wrong, only3% more airline passengers considered luggage handling as one of the mostimportant factors when choosing their travel mode.

Figure 3 – Differences in gender.

A similar feature is identified comparing English and French customers,however in this case the difference in sensitivity to the ticket price is muchmore significant than in previous examples. Almost 20% more UK thanFrench respondents base their choice of transport mode on the price of theticket. Unlike the French customers that seem to be more sensitive to theaccess to the airport, see Figure 4.

Advanced Concepts


Figure 4 – Differences between French and English passengers.

Another counter-intuitive result was discovered comparing frequent travellers(travelled more than 6 times a year) and not frequent travellers (less than 6times). In case of ‘not frequent’ travellers ticket price showed to be the mostimportant attribute that significantly influences passenger modal choice, muchmore than in the case of frequent travellers (as seen in Figure 5).

Figure 5 – Differences between frequent and not frequent passengerstraveling by rail.

We assumed that frequent travellers should be more sensitive to price sincethey spend a larger amount of total budget on travelling than passengerstravelling only on several occasions.



On the contrary it was the schedule & frequency and on-board comfort thatdecided frequent passengers’ inclination towards certain mode. Presumablythis feature is due to the fact that most of the frequent travellers are notpaying for their travel ticket since the majority of frequent travellers aretravelling for business (see Figure 2).

8. Air transport passengers

Unlike in case of Thalys and Eurostar passengers, between Roissy CDG andLisbon Airport passengers there was a very slight difference in travelpreferences. This similarity is due to the fact that airline business is a veryhomogeneous world offering often identical services (meal on plane,newspaper, frequent flier points, friendly attendants), while often using thesame type of aircraft on medium-haul trips. Figure 6 outlines some majordifferences in air and rail passengers’ preferences; however a further analysis isneeded to discover possible relationship in passengers’ travel choices.

Figure 6 – Differences between air and rail passengers.

Both air and rail passengers agree that the most influential factors are ticketprice and travel time. Although they seem to strongly disagree about theimportance of the access to the airport or train station, walking & waiting timeand schedule & frequency. As already mentioned luggage handling proved tobe less important as we assumed in the beginning of the survey andsurprisingly on-board services had no or very low impact on passengers’ travelchoices.

Advanced Concepts


9. General behaviours

In order to learn more about passenger behaviours we have listed a series ofevents related to travelling. We used the verbal frequency scale that indicateshow often an action has been taken. The questionnaire has revealed manyinteresting findings, see Figure 7 below.

only 7% of respondents would be always willing to pay more for a trainticket than for the flight ticket, what might prove to be a burden especiallyin case of Eurostar where the rail ticket is often more expensive than airticket ;

64% travellers find connection issues always and often very important, andthe same percentages of travellers are used to waiting at airports longerthan 1 hour before the time of their scheduled flight ;

opposed to that 55% of respondents arrive at the train station less than _hour before their scheduled train departure time ;

as much as 66% always and often base their choice of transport mode ontotal travel time ;

34% strongly deny willingness to spend money for luggage check-in at arailway station (service exists at Leipzig-Hale airport) ;

35% of travellers find connection issues always very important and 21% ofrespondents would pay extra charge for luggage to be delivered to theirdomicile, after they conducted their multimodal journey.

10. Decision Tree

A different approach was applied in order to evaluate probabilities ofpassengers choosing a certain transport mode. A decision tree takes as inputan object or situation described by a set of properties, and outputs a yes/nodecision. Decision trees therefore represent binary functions. A decision treeis an arrangement of tests that prescribes an appropriate test at every step inan analysis. More specifically, decision trees classify instances by sorting themdown the tree from the root node to some leaf node, which provides theclassification of the instance. Each node in the tree specifies a test of someattribute of the instance, and each branch descending from that nodecorresponds to one of the possible values for this attribute. The tree is atrade off between the highest possible percentage and the number of splits.Naturally it is possible to predict 100% probability of a passenger saying yesor no to a certain choice but it would result into great number of splits.



Figure 7 – Decision tree for HST choice.

The data input of 914 answers was used to predict the probability ofpassengers saying ‘YES’ to a choice of high-speed train (see Figure 7). Theway to interpret the tree is the following. Take an example of the rightbranch. 510 rail passengers (95% of all rail passengers asked) agree thatluggage handling is not important for them. With probability of 91% thesepassengers would choose HST according to the tree. Based on the examplethere is a significant difference in air and rail passenger preferences; the firstinformation computer considered at root node was the transport modepassengers were traveling with. Further analyses are required to find the mostreliable tree, regression analyses will be used to test the significance of eachinput variable.

11. Demand model

We all know that models are only simplified representations of reality whichcan be used to explore the consequences of particular policies or strategies.When modelling choice made by passengers we need to take intoconsiderations several variables. Logit models have been widely used in the

Advanced Concepts


analyses of the choice. The most common variables used in the modelsanalysed before are: travel time

travel price

schedule & frequency

travel purpose

possibility of using another mode (car)

direct vs. indirect routing generalized time of travel

weighted measure of air flight time

measure of the need to interchange

Numbers of analyses done in the past suggests that both logit (mode choice)and an econometric (regression) approach are valid for estimating the effectof the high-speed rail on air demand. The choice depends on the availabilityof data. In order to implement the results of the questionnaire into the modelwe will use logit model of individual choice. This model uses the relativeattractiveness of each of a set of options to predict, for each option, theprobability of an individual choosing it.

12. Future work

To simulate a European transport network and possible modal split we willneed to take into account passenger behaviour, existing and forecasted high-speed train infrastructure and among many other things the situation in airtraffic in relation to congested airspace and airports. Based on recentexamples in Europe we assume that there is a high possibility in achievingsignificant en-route and airport capacity improvements, while satisfyingpassengers needs at the same time. Thanks to inter-modal transport somecongested hub airports will be able to free as much as 10% of their runwaycapacity. In Spain, the replacement of Madrid/Barcelona andValence/Barcelona services by HST could free up to 19% of the runway slotsat Barcelona (data source: Network, Capacity and Demand Planning,EUROCONTROL Experimental Centre).

However the future evolution of integrated transport networks will mostlikely depend on the airlines willingness to co-operate with railway operators.Examples show that some airlines will prefer to maintain air services oncertain city-pairs (Madrid-Barcelona with 64 flights a day) while competinghead to head with railway operators. In order to keep up with competitionand attract more passengers airlines will need to operate smaller aircraft withhigher frequency; resulting in more aircraft flying in the European sky each



with less seats on-board. Needless to say this kind of outcome will put morepressure on air traffic services and create additional problems in the future.Hence inter-modality will have an inverse effect, causing more congestion anddelay.

In order to better understand the impact of inter-modal transport on airtraffic different scenarios will have to be considered. The success of inter-modality and hopefully the possibility of easing congestion will depend onpassengers’ willingness to experience new way of travelling, operators’willingness to co-operate and most of all the influence of low-cost airlines andtheir future evolution.

13. Conclusion

Passenger travel preferences vary depending on nationality, purpose of traveland even age. Based on questionnaire results we can conclude that it is verychallenging to find a common denominator that can describe in details themost significant transport factors. However 60% of the passengers agreedthat travel time, ticket price and access to the airport or station are the threemain decision factors. About 40% of all passengers consider on-boardcomfort, schedule & frequency, punctuality & reliability and walking/waitingtime crucial when choosing between different transports modes, while on-board services and luggage handling have little or no influence on passengers.Some factors proved to be important only to a small group of people,although they can be very influential; fear of flying or fear of crossing theChannel Tunnel (Eurostar) will certainly strongly influence passengers’inclination towards the competitive mode.

After opening Eurostar services the behaviour of French travellers have notchanged significantly, opposed to English customers that formed most of thetrain population. Our questionnaire has resolved that UK passengers aremuch more price sensitive than French customers. Similar differences inbehaviour have been discovered when comparing men with women andfrequent with not-frequent travellers. No matter how big the differences are,price and time proved its priority in every comparison.

14. References

[1] EUROCONTROL PRU Performance Review Report # 7, PerformanceReview Commission 2004, Brussels.

[2] Cokasova, A., (2003). Modelling of Air Rail Intermodal Transport at MajorEuropean Airports from the Passenger Perspective. Final Thesis, Universityof Zilina & EUROCONTROL Experimental Centre, April 2003, France.

Advanced Concepts


[3] Rietveld, P. & Brons, M. (2001). Quality of hub& spoke networks; the effectof timetable co-ordination on waiting time and rescheduling time. Journal ofAir Transport Management 7 (2001), p. 241-249

[4] European Commission -European Cooperation in the field of Scientificand Technical Research (1996). Cost 308. Interaction between HighSpeed and Air passenger Transport, April1996.

[5] Ellwanger, G. (2002). Successes for high-speed rail. Rail International,International Union of Railways UIC, September 2002, Brussels.

[6] EUROCONTROL Experimental Centre, Network Capacity andDemand Planning, Study of Low-Cost Traffic Patterns, March 2004.

[7] Alreck P., Settle R., The Survey Research Handbook, Guidelines andStrategies for Conducting Survey, Maryland 1995.

The Airport of the Future or Breaking the Constraints between the Terminal and the Runways


THE AIRPORT OF THE FUTURE OR BREAKING THE

CONSTRAINTS BETWEEN THE TERMINAL AND THE RUNWAYS

Marc Brochard

EUROCONTROL Experimental Centre, Brétigny, France.

From EC 6FP "Aeronautics and Space" Work Programme [1], itself based onthe ACARE vision 2020: world aeronautics is entering a new age of aviation,the age of sustainable growth, characterized by the need of more affordable,cleaner, quieter, safer and more secure air travel. European aeronautics iscommitted to play a prime role in shaping aviation for this new age. Researchand technology development will be essential in responding to this challenge.

The 6FP aeronautics research work programme of the thematic priority“Aeronautics and Space” is planned in accordance with the relevant part ofthe Specific Programme ‘Integrating and Strengthening the European ResearchArea’ and with the Strategic Research Agenda1 prepared by the AdvisoryCouncil for Aeronautics Research in Europe (ACARE) [2]. The StrategicResearch Agenda (SRA) has set out the directions for European research inthe next decades towards fulfilling the ambition for the future of aeronauticsestablished in the Report “European Aeronautics – a Vision for 2020”, as wellas in the White Paper ‘European transport policy for 2010: time to decide’,adopted by the Commission in September 2001 [3].

From the Challenge of Air Transport System Efficiency as defined in theACARE “Strategic Research Agenda” [2]: the objectives laid down in theVision 2020 report are extremely ambitious for European Air Transport.They, and other figures, foresee a tripling of traffic in Europe, both in terms ofpassengers and flights, with a punctuality target of 99% of flights departing andarriving within 15 minutes of their timetable in all weather conditions.Passengers should not have to spend more than 15 minutes in the airportbefore departure and after arrival for short haul flights, and 30 minutes forlong haul flights.

The challenge set was how to achieve these objectives through investigationsinto the Future Air Transportation Environment, the aircraft as part of the AirTraffic Management (ATM) infrastructure, Airports and security issues. Themain conclusion of the work undertaken in the different fields was that theVision 2020 objectives would not be reached unless there was a paradigmshift in the way the Air Transport System is conceived and operated. The

Advanced Concepts


Research and Development (R&D) paths, which pave the way towards thesechanges, are amongst several the following: maximize current airportperformances and the Airport of the Future.

1. The Airport of today

In light with the airport of the future and also in order to support theforeseen traffic growth over the coming years, several limitations andconstraints at the airport level that require urgent need for evolutions, havebeen already identified in the ACARE SRA [2]:

Security is a key issue for aviation security and will require improvement inthe future. Unless it is a key process, it is also seen as limiting factor toaccommodate with the coming traffic growth: security process takes toolong. This process shall be re-organised to achieve speedy and efficienthandling.

Safety shall not be reduced when accommodating with the traffic growth.Today, the air transport infrastructure capacity is directly constrained bysafety objectives. This requires paying great attention to safety issues whenhandling more aircraft at the airport side.

Environment is becoming a strong limitation for airport creation orextension. There is a strong need to reduce as much as possibleenvironment impact (such as gas and noise emission) so that it becomesacceptable by the population leaving in the vicinity of the airport, and alsoto meet the overall world wide objective to reduce pollution.

Institutional and economics issues shall not be forgotten. Any evolution orinnovation at the airport side shall still allow the “users” and anystakeholders to develop their own business according to their ownpotential.

Runway performance is a physical limitation today: the current aircraft andground facilities (landing aids mainly) performances do not enable toincrease the number of aircraft landing or taking off. Also, due to otherlimitations (like environment or available ground surfaces), it is verydifficult to create new runways. Therefore, runway performances andoperation shall be revisited to accommodate with the traffic growth.

Passenger and luggage process needs to be dramatically changed even if itdoes represent a part only of the complex multi-factor process at theairport. But, it is well recognised that the performance and efficiency ofthe over all Air Transport System heavily rely on the ability to processpassengers and luggage in a short time frame.

Weather (especially bad weather conditions) is always reducing airportcapacity. This shall not be anymore a limitation for tomorrow.



On top of the limitations already well identified by the SRA, there are otherlimiting and constraining factors that shall not be forgotten and which need tobe considered and revisited when envisioning the airport of the future. Thefirst one is about the airport location. Today, mainly for historical reasons,many airports are located very close to the main cities and thus very close tothe passenger. This location in the vicinity of areas with high density ofpopulation constraints any evolution because large surfaces may not beavailable for potential extension, and because of environment issues (likenoise and gas pollution) which are getting more and more sensitive for thepopulation. Nevertheless, it shall be noticed that few airports have beenrecently build far away from the city. But there is common shared belief thatairport shall not too far from the city. The second concern is about theairport infrastructure which in all cases today, is a complete integratedinfrastructure providing more and more non ATM services like for exampleparking areas, access public ground transport means, leisure facilities (such asrestaurant or duty free shop), and ATM services like boarding gates and flightoperation (tarmac, taxiway and runways). This implies to build very largeinfrastructures with lot of people working permanently in the same area, withpeople having no direct link with Air Transport business. The last one is moreabout air transport services, meaning the set of destinations offered from theairport. Each main city has one, when not several, airport infrastructure(s)aiming at offering the maximum of destinations, be it short, medium or longrange. The aim is to offer the largest service to the passengers. But, this resultsof having a very complex air route network (lot of city pairs, lot ofconnexions and lot of crossing axis), which makes more and more difficult theco-ordination of the overall ATM partners and the management of the trafficflows.

2. New paradigm for the Airport of the Future

Many proposals are already been discussed, many researches are ongoing todraw the airport for the future: underground terminal facilities, offshoreairport site for example. There are also many solutions that have been alreadyput in operation mainly those aiming at developing the inter-modality oftransportation mean by linking the rail with the flights. But we could alsoadopt a new fresh perspective based on the following paradigm: breaking theconstraints between the terminal – the airport land side – and the runways –the airport air side. – meaning putting the airport terminal area far away fromthe gates and runways areas

Advanced Concepts


Applying this new paradigmmay enable to put the airportland side – the terminal area –back to the cities with a betterintegration with other terminalareas such as rail way stations.Also, we could imagine that theairport air side – boarding gatesand runways – could bedeveloped far away the citiesand connected via dedicatedhigh speed transport means toone or several terminal areas.Going further, we could imagineeither to develop large runwaysareas connected to severalcities (in other words, sharing airport and thus reducing the overall number ofairports), or imagine to distribute runways in a different way than today.Several scenarios may be explored. In any cases, this could enable a betterdistribution between air and ground transportation means.

3. Impact for airport landside

Breaking the constraints between terminal and runway would result in puttingthe terminal area – airport land site – including inter-connection with othertransport means, parking area, leisure facilities, back to the city. The airportterminal facilities will be closer than today (or perceived as such) to thepassengers and workers. This should reduce the number of people workingon the airport air side with probably positive impact on airport security: onlyworkers operating flights will be on the air side. Impact on the environmentwould be less penalising as the aircrafts and runways will be far away from thecities in area with low density of habitants. On the other hand, we couldimagine better balancing the migration flux using dedicated transportationmeans between the land and air sides. For example, fewer workers on theairside will result in less people travelling from their home to the airside areas.Other worker operating at the land side, the terminal area, will stay in city.

Passenger and luggage process would be improved, as it will not beconcentrated as today. As passenger will have to use ground facilities beforegetting into the aircraft, this will give more time for security check of luggageand passenger. We can imagine using that period of time to process theluggage when passengers are moving on the ground from the terminal area tothe boarding area. This would enable to provide a “with no delay boarding”


Future airport paradigm



phase (at least perceived as such by passenger). Same process may beperformed at the arrival: i.e. using the ground travel period of time toredistribute luggage to the passengers. Again, the perception will be no moretime at the arrival to get the luggage back. There would be no more wastingtime waiting in the airport as today.

Passenger will use dedicated high-speed transport means to go from theterminal area to the airport gates (the aircraft) and vice versa. The tripduration shall be less than an acceptable duration (for example less than 30’).We could imagine automating the boarding meaning that the passenger willgo into the aircraft without having to change vehicle and vice versa at thearrival. Even further, we can imagine breaking up the aircraft itself: aircraftcould be composed of fixed airframe with several vehicles able to move onthe ground and which could be plugged inside the airframe for the flightphase. This could be seen as a docking station processing, station being theground vehicles and docking being the airframe.

By increasing inter-modality of transport means, we could image to see apositive impact on the overall cost of air transport (compared as what it coststoday). Travel cost would be spread over air and ground transport means.Finally, impact on the environment should be positive, as we will operate flightwhere it is really cost effective to do so. Some part of the travel will beperformed by ground high-speed means. We assume that tomorrow, air andrail transport will not be competing, but rather seeking to complementary andoffering new transport business where passenger gets a unique ticket fortravelling from A to B without caring about the transport means used, be it acombination of rail and air transport means or not.

4. Impact for airport airside

Boarding gates and runway would be concentrated far away from the city(high density of population area). These facilities could be “shared” betweenseveral main cities which could result in a reduction of the total number ofairports (reduction of city pairs and runways concentration). We could alsoimagine instead of having concentration of runway on one unique large area,to distribute smaller set of runways around the city(ies), theses smaller set ofrunways being “specialised” for given traffic.

Advanced Concepts


Large runway set sharedbetween 2 cities.

Runways distributed around city.

This could also reduce the complexity of the air route network. In a verysimple way, we could imagine that fewer airports will reduce the number ofcity pair (in fact, it would be better to speak about airports pairs) as shown inthe examples below with a theoretical network of 20 points with a completepoint to point connexion on the left side, and on the right side, the samenetwork where few points are grouped together (4 pairs and 1 triple). Thisshould enable easier or more efficient co-ordination with the entire airspaceusers and ATM community.

Complex point to point networkwith 20 points.

Same network with only 12 points.

As number of new airports would be reduced; we can imagineconcentrating/integrating various operation tasks on the single area (wheretoday it is dispatched over multiple areas). This should provide reduction inthe operation cost and be beneficial for the overall air transport businesscosts. This might not be the case with the scenario of redistributing therunways around the city(ies)



Being far away from the cities, environment constraints, even if, still, it shall beconsidered, will be less penalising for airport infrastructure and shall have lessimpact on the vicinity (mainly the citizen).

If we imagine automating the boarding (like docking station) and if weconsider the de-localisation of the terminal areas, then we could imaginereducing, if not removing the taxiway. Aircraft boarding might be locatedcloser to the runway as compared as today. At least, the complexity of thetaxiway network should be reduced and thus having positive impact onoperation cost and ground traffic control tasks.

Number of runway would be greater than today due to the reduction ofenvironment constraints and also the ability of using larger ground areas. Thisshould have a direct positive impact on runway/airport performances.

Fewer workers will be present on the airport side (connection with othertransport means and leisure facilities will be far away). This should havepositive impact on the security of the site and the flights.

5. Exploration threads

This “Airport of the future” needs to be explored, developed and validatedkeeping in mind several main threads that shall not be under estimated even ifit goes well beyond the pure ATM world. The list is not a complete list but itaims at paving the way for further work. All these analyses should be basedon a door to door concept meaning for the passenger that I start when Iwant and from where I live. An other strong assumption to consider is theneed always for the passenger to reduce if not to remove any waiting time(remember ACARE statement about passenger time spent in the airport: lessthan 15’ for short and medium haul, and 30’ for long haul).

Transport network (ATM and inter-modality) analysis: the aim is to analysethe combination of various transportation means and to analyse the impact ofbreaking the constraints between land and air sides such as current andforeseen evolution of the ATM Network in Europe; current and foreseenevolution of the Rail network in Europe (other ground transport means shallbe considered); mapping ATM and Rail network (current and foreseen) toidentify weaknesses if any; impact of city pair evolution on the ATM network(reduction of complexity, capacity improvement). Several scenarios shall beanalysed such as reduction of city pair by grouping airports together (severalcities may use the same runway platform), or dispatching runways around city.

Airport infrastructure and integration with the overall ATM: the aim is toevaluate how the new airport infrastructure would look like including securityand safety issues; and also to evaluate how the new paradigm would beintegrated with other infrastructure such as the ATM. Several investigations

Advanced Concepts


might be undertaken such as landside integration with other terminal areas(railway station); security analysis for both land and air side; impact of change,challenge, solutions and connection between land and air sides. Other workcould be focussing on new concept for runway use. For example, as there willbe less direct connection to be performed by flight between land and air side(boarding being automated), is there any need for taxiway? Having morerunways does it lead to new runway procedures? Same question might beasked for the airways. We could imagine to see a development of “navette”(a flight every hour, if not more) which could lead to think about formationflights. Last but not least, ATM integration between terminal and en routeareas should be investigated also. Strong investigations shall be performed inthe field of impact of this new paradigm on ATM tasks and responsibilities.We could imagine having room for automation.

Institutional and economics: the aim here is to analyse any impact of breakingthe constraints between land and air sides, on the overall airport andtransportation business, including the assessment of the real expectation ofthe final client. It is also needed to pay great attention to institutional aspectunderlying the transportation business. Final client shall be clearly identifiedand his/her needs really assessed. Analysing the overall transportationexpectations shall be analysed considering the various kind of the airtransport: tourism versus business transport versus fret. Also, there will be aneed to assess the current stakeholders for both air and groundtransportation means; and to analyse the impact on these businesses and forthese stakeholders of breaking the airport in 2 parts (some might win, somemight lose, new stakeholders and/or businesses model might appear). Politicalimpact shall not be underestimated especially if we consider a reduction incity pairs, meaning reduction of large airport; and also analysing the impact ofsharing airport infrastructure between different regions.

Environment: the aim is to evaluate any environmental impact, would it bepositive or negative bearing in mind that environment may constraint a lotany evolution of the airport.

Transition plan: the aim is to check how realistic the application of the newairport paradigm is: identifying area where new airport might be possibleversus small airport where current infrastructures may remain as today in thefuture (do we need to change all existing airports?), elaborating transition planfrom current infrastructures and what could be the airport for tomorrowincluding the possibility of breaking existing airports, and cost benefit analysisto assess the economical value of this new concept.

Even if the complete application of this new concept might be seen as pureutopia, we do believe that working in that direction would lead us to better



understand the airport business and to think about other means to improveairport efficiency, this being the ultimate goal of our work.

6. References

[1] EC 6FP "Aeronautics and Space" http://www.cordis.lu/fp6/aerospace.htm

[2] ACARE "vision for 2020" http://www.acare4europe.org/

[3] EC, White Paper “European transport policy for 2010: time to decide”,September 2001.

The Airport of the Future or what can be the Airport in the Year 2020 and After ?


THE AIRPORT OF THE FUTURE OR WHAT CAN BE THE

AIRPORT IN THE YEAR 2020 AND AFTER ?

Martin Matas(1,2), Marc Brochard(1)


(2) University of Zilina, Slovakia.

According to current and forecast growth of air traffic in Europe it isexpected that the air traffic will double or triple within next 20 years in termsof both aircraft and passengers. Many major airports have already today theproblem with airport capacity and the runway and terminal development isconstrained by lack of the space. But there are still many possibilities how toincrease airport capacity or airport performance. Some of them initially seemto be less safe. Safety shall not be reduced when accommodating with thetraffic growth. Today, the air transport infrastructure capacity is directlyconstrained by safety objectives. This requires paying great attention to safetyissues when handling more aircraft at the airport side.

Tripling of the air traffic requires a considerably different approach in theairport concept. The purpose of my thesis is to propose new concepts of thefuture airport which will be able to meet the air traffic demand in the yearsaround 2020 and after.

1. Initial ideas

The first research axis aims at lookingfor solution for breaking theconstraints between the terminal andthe runways. The main idea is to placethe landside terminal in the city and tobuild new airside far from the citywhile connecting them with HighSpeed Train. There are many positivesand as well some negatives of thisconcept. Only a few are mentionedhere. For the cons of this concept itwill be necessary to find solutionswhich will mitigate the negativeimpacts.


Future airport paradigm

Advanced Concepts


The positive aspects of this concept might be the following:

one runway system shared among several cities helps to reduce thenumber of airports ;

from the passenger point of view the terminal will be closer than today ;

there might be more than one terminal in the city which would decreasethe travelling distance to the terminal ;

there might be a security check in the train which will save time ;

immediate boarding after arrival to airside will support no-delaydepartures ;

the total travel cost will be spread over air and ground transport means ;

fewer workers will be present on the airport side which might havepositive impact on the security of the site and the flights ;

much less people would be affected by pollution (gases, noise..), becausethe airside would be far from densely populated areas ;

there will be much more space for runways, taxiways and aprons, whichcan have positive impact on safety ;

passengers will be gathered longer period of time before the take off andthis will enable better management and estimation of aircraft departure -door to door concept.

The negative aspects of this concept might be the following:

high initial costs for high speed train and it may be difficult to fund thisexpensive project ;

airport's business may be affected by replacement of the shops,restaurants, entertainment, parking, etc. to the city landside ;

airside part of employees will have to travel a significantly longer distance ;

airport far from the city will lead to relocation of the destinations flown byairlines caused by increased inter-modality between air and rail transport ;

most of the passengers would have to change transport vehicle one moretime.

The second main research axis aims at envisioning radically different approachto a traffic flow at an airport. There are many particular capacity constraintswithin an airport. Each of them offers the possibility for improvement. Theremay be a radically different approach to the aircraft flow at an airport whichmay speed up the process significantly to meet future demand.



Capacity constraints sequence.

Terminal capacity may be solved by building underground or multi-floorterminals. To release the airport from terminal with number of passengersand facilities, placing landside terminal to the city seems to be a good solution.Reduction of the time spent by a passenger within the terminal can bereached by speeding up the check-in process - docking station (SRA 2)

Apron capacity may be increased by speeding up the turnaround process.While many airlines have turnaround time about one hour, there are somewhich can turnaround within 30 minutes or less with the same type of aircraft.Other option for significant increase of the apron capacity is building an apronwith two levels.

Runway capacity is constrained either by runway occupancy time or by wake-vortex separations by ATC. Runway occupancy time may be significantlydecreased probably from 50 seconds down to 30 seconds if accurate rapidexit taxiways are applied and if new runway procedures for lining up, takingoff and landing take place. For quicker exiting the runway there might be a

Advanced Concepts


kind of turning runway ortaxiway, which wil lenable the aircraft safelyturn off from the landingdirection at higherspeeds while otheraircraft would alreadystart the take off run. The idea is optimally redesign the traditional shape of arunway.

The taxi time is sometimes high due to the complexity of the apron, taxiwayand apron system. Reduction of the taxi time might contribute to the overallsave of time. Generally many possible solutions may come up within theresearch and each solution will be investigated and compared among theothers.

A third research axis aims at reviewing the design of an airport andenvisioning multilevel runway, taxiway and apron system. Many of currentairports can not build another runway next to the others because of airportborder with the city. Can you imagine that there is another level of runwaybuilt over the other one, built over the terminal and apron or crossing theother runways like a bridge? These kinds of solutions might break the currentspace constraint on many airports. This idea also motivates the existence ofmultilevel taxiway and apron system with multilevel terminal and therefore itleads to direct increase of capacity of the whole airport.

LHR airport with two level runwayand apron.

Two level configuration with parallelrunway over a taxiway.

There are many configurations available. The upper runway doesn’t need tobe right over the other one. It may stand in parallel to other runway enablingaircraft taxiing under it to the lower runway, it can cross the lower runway inthe middle or it can stand completely apart from any runway for example onthe lower apron and lower terminal.



2. Conclusions

After some investigation on the current projects dealing with the airport ofthe future, there is huge variety of solutions and ways of improving/ reviewingthe airport and its capacities in order to cope with the challenges of the15/20 years to come. In order to look for innovative airport that shouldsupport sustainable Air Transport business development, focus will be givento the concept of breaking the constraints between the terminal and therunways.

3. References

[1] http://www.airliners.net.

[2] Airport Operations / Norman Ashford, H.P.Martin Stanton, CliftonA.Moore. - 2nd ed.

[3] ACARE "Vision for 2020" http://www.acare4europe.org/.

[4] Airport of the Future, Marc Brochard, Eurocontrol Exp. Centre, edition2004.

Project « Paradigm SHIFT »


PROJECT "PARADIGM SHIFT"

Laurent Guichard(1), Sandrine Guibert(1), Horst Hering(1),Jean Nobel(2), Didier Dohy(2), Jean-Yves Grau(2),

Khaled Belahcène(3),Marc Brochard(1)

(1) EUROCONTROL Experimental Center, Brétigny, France

(2) STERIA

(3) CS Systèmes d’Information, Clamart, France.

There is a common agreement that current air traffic management systemcannot cope with the challenges of future air transport system (ACARE SRA-I, 2002; University Concept team, 2003; EUROCONTROL OCD, 2004; Gateto Gate project, 2004), and a new control paradigm seems to be inevitable.The project Paradigm SHIFT, sometimes called SHIFT, started at the beginningof 2004 at EUROCONTROL Experimental Centre, has been one attempt toresponse to this need. SHIFT goal has been to investigate a new controlparadigm that could cope with future air traffic demand of the horizon 2020and beyond. A group of several experts coming from ATM domain,operational air traffic controllers (ATCo), human factor, layer and systemarchitecture went through several brainstorming sessions and finally came upwith five main interrelated operational concepts: Contract of Objectives,Operational Plan, Target Windows, Decentralized Airspace, and DualAirspace. The analysis introduces a new way of designing the air navigationinfrastructure based on the concept of management by objective instead ofby means. It defines the foundation of an ATM system able to cope withtraffic demand at the horizon of 2020 and beyond, while maintaining a veryhigh target level of safety and supporting a sustainable air transport businessdevelopment.

This group of expert has also elaborated an innovative approach for newATM concept definition and validation via a research agenda listing any issueswhich shall be considered when designing a new concept: the researchagenda which is becoming the backbone of any INO work in the field ofadvanced concept.

1. Current ATM characterization

In order to define a future air transport system, the first action was tobaseline the current one, meaning identifying the key features of it and

Advanced Concepts


defining which constraints to be considered when designing the new concept.Starting from the results of the Super-Sector project [12] and severalinterviews with ATCo and ATM experts, the following conclusions weredrawn:

The En-route air traffic is a mix of climbing, descending and cruisingaircrafts. On one hand, each of these categories has differentcharacteristics in terms of density, disruptions, bulk, shape, complexity, andservices, which require different resolution approaches. On the otherhand, the sharing of tasks and responsibility among ATM actors has alwaysbeen based on localized, geographical sectors where the control of alltraffic categories is unique. Thus future concept shall consider the differenttraffic characteristics and resolution approaches in the design of airspacethrough a well-defined task sharing between actors of ATM.

ATM is inherently stochastic because all its components are continuouslysubjected to disruptions. Indeed, in current ATM paradigm, uncertainty isone major factor that could characterize the consequences of trafficgrowth, or at another level of description: delays, and could be seen asthe results of disruptive events on the deterministic planning components,caused by stochastic nature of ATM. Disruptions can be classified intodifferent categories:

ad-hoc events e.g., those caused by unexpected weather,runway conditions, aircraft failure, etc.

inherent events or constant imprecision such as inaccuracy ofplanning mechanism, technology, models;

and network events or system-wide problems generated byinterfaces between ATM components e.g., ATFM vs. ATC, andATC vs. aircraft crew.

The future ATM system shall not try to eliminate these uncertainties butmust appropriately control and monitor them. The management ofuncertainties is indeed one key issue for future ATM.

Aircraft operating cycle shall be seen as operational continuity betweenland/ground and air operations. In this view, not only take-off and landingtime are crucial for airlines but also all ground and air operations.Therefore the operating cycle of an aircraft, called rotations for airlines,must be integrated to the whole ATM system, from planning tooperations on Airside and groundside (or landside).



Figure 1– Operating cycle of an aircraft.

On top of these three main characteristics of the current ATM, a commonconsensus came up stressing the need to have a holistic approach for theATM operation based on three fundamental elements: Traffic organization,airspace structure, and control procedures and practices. It was clear in thegroup of expert that any proposal for a new traffic organization must takeinto consideration the impactson the procedures andairspace design, and must notbe solved by a globalapproach . The s t rongrelationships between thethree elements are describedas the air navigation tripod andshow that only a local level ofexpertise can produce theright balance between axes.

2. The Paradigm SHIFT Concepts

By taking into account the key features of ATM mentioned above, we suggesttwo majors’ paradigms as the backbone for the shift of control paradigm fromthe current situation to the future: Contract of Objectives being a proposedway to cope with the objective of arriving on time, instead of departing ontime as today; and Dual Airspace being a proposal for reviewing the airspaceinfrastructure to provided are supporting high density of traffic with highcapacity. These 2 fundamental concepts have lead to define 3 other mainconcepts which are the operational plan, the target widow and thedecentralized concept.

Advanced Concepts


Main SHIFT’s paradigms.

These paradigms could be independent but could also be combined togetherbecause there’s no contradiction in their mode of operations.

The Contract of Objectives defines objectives applicable to flights and linksATM actors together through agreed interfaces. This Contract of Objectives isdrafted during a negotiation phase involving all actors i.e. airlines, airports,ANSP, military units, etc. whereas individual objectives are assigned throughthe breakdown structure of the responsibilities locally at the level of thecontrol centre. We assume that local actors have the best view of how tooptimize their organization. By doing so, at each local organization, therewould be a decentralized ATM structure. The objective assignment andnegotiation can be performed as a collaborative decision-making process toestablish the Operational Agreement.

The so-talked Operational Plan doesn’t preclude any strong constraint on theagreed objectives, but a certain degree of tolerance is used instead.Disruptions are part of the ATM system. Putting constraints to insure safetyand fluidity is necessary to manage the traffic. But over-constraining theobjectives close the door to the necessary resilience to deal with uncertainty.The notion of Target Windows is suggested here to define the intermediateobjectives for a flight, where a target is associated with an interval called 4D-windows. These target windows are supposed to be the negotiation tool orthe mean to achieve the Operational Agreement.

The paradigm of Dual Airspace introduces a small number of continentalhighways conveying long haul cruise traffic in complement to all-includedsector-based traffic (district) as of today. This purpose of this paradigm is torelease the pressure on local air navigation services by separating the long-haul routes from the current routes. Long-haul routes could be assimilated tohighways while local routes are national routes.



3. The Contract of Objectives

Air navigation service efficiency requires better functional and operationalcontinuity between the various actors, whether they are air traffic actors orthose playing a role in the more global air transport system (airlines andairports). There must be an operational link between all these actorsidentifying the role and the resulting redistribution of tasks and responsibilitiesfor each actor, in relation to a clear, well-defined objective that is accepted byall concerned. This objective is general, of course, and will be different foreach actor in accordance with the actor's specific characteristics andworkload. The challenge is to define a common operational minimum amongthe actors which is sufficient to strike the right balance between productivityand safety. For this reason, it is helpful to propose a global contract for the"air" segment of the aircraft's operating cycle. Firstly, this would facilitatefunctional and operational continuity within the ground segment, since it iscompatible with the objectives of airports. Secondly, it would play a role inintegrating the flight segment into the rest of the system, by creating bonds ofreciprocal responsibility between the airlines, the aircrews and air trafficactors. The proposed name of this contract is the contract of objectives. Thecontract of objectives is associated with one flight. The contract of objectivesis intended first of all as a guarantee of results offered to the airline by the airtraffic system on the basis of known constraints at the time when the contractis drawn up. Consequently, it is the ATC responsibility to fulfil the contractonce all actors accepted this one. For controllers, the incorporation of thecontract of objectives into their activities brings an additional task. It is clearthat respecting the contract of objectives becomes a key priority in theiractivities, safety remaining the controller's top priority. If the contract ofobjectives cannot carry out during the flight, it is renegotiated at strategic levelin the operational plan process. The "contract of objectives" is not a rigidframework within which aircraft have to operate. It contains built-in marginsfor flexibility and adjustment in order to manage disruptive factors. Thesemargins are compatible with those of the other components of theaeronautical system. The contract of objectives is therefore a flight envelopedefined on the basis of:

The aircraft's room-for-manoeuvre ("commercial" flight envelope). The predictions relating to en-route control constraints.

The final objective to be attained (i.e. destination punctuality). The closerone comes to the final objective, the smaller the room for manoeuvrebecomes.

Like for the controllers, the "contract of objectives" significantly alters the roleof aircrew in the conduct of the flight. They are no longer the only persons

Advanced Concepts


responsible for adhering to the arrival time at the destination. They cannot, ofcourse, question the "contract of objective" once it has been accepted by allthe partners. As long as the flight takes place within the envelope defined inthe contract, it falls to the controller to give orders to aircrews regardingsafety and navigation. It goes without saying that under no circumstances cancontrollers pilot the aircraft. All orders from controllers are submitted forapproval to and executed by the aircrew. This means that the aircrew has atits disposal on board the aircraft information telling it the "adherence" of theaircraft in the contract of objectives.

4. The Operational plan

The process by which the contracts of objectives of all flights (i.e. refineddemand) are elaborated is the operational plan. The operational planmechanism is both a negotiation and refinement process between all actorsinvolved in the air operations (airlines – airports – ATM/ATC providers).“Refining is more efficient then redefining”. This method permits to optimizeco-operations and allows agreement to be found in an early state of theprocess. In this case, refinement appears as implicit agreements becauseelements are specifying more in detail and not “called in question”. Thisphilosophy drives the whole process of the planning phase of the ATM. Theprocess of drafting "contracts of objectives" is at the heart of the air transportsystem, since this process will define the framework within which flights willbe performed and the responsibilities which will be applied to ATC actors.The contract of objectives is drafted on the basis of two sets of requirements:

The individual requirements of the flight in question.

The general or global requirements of the air transport system and all itspartners. The individual requirements are a subset of the generalrequirements.

The aim of the operational plan is to better manage the scarce resourcesrepresented by runway capacities and ATC bottlenecks. To this end, the aimsof this approach to the drafting process will be:

To adjust the resources available to fit demand. This adjustment is a two-way process, i.e. ATC resources are adjusted in accordance with userdemands in the full knowledge that the resources are limited and will notbe able fully to satisfy demand. This also constitutes an acknowledgementthat for certain areas of airspace, it may not be possible to satisfy thewhole of the demand. The system will, however, be optimized in order tosatisfy demand as far as possible.

To enhance cooperation between the various actors in air transport inorder to share and work on the most precise and up-to-date information.



To minimize and/or attenuate global problems in order to encourageadjustments and limit the drawbacks.

To reason at each stage of the drafting process with an appropriate levelof granularity that depends on the precise information and the timeremaining for the issuing of the final contract.

To use "real-time" information as soon as it becomes available in order toincrease the precision of the planning.

Operational Plan process.

The challenge at the heart of the drafting process is therefore to build asystem based on adaptive procedures and the sharing of considerableamounts of information. Operational plan is a three-step process leading tovarious releases of the global agreement (i.e. operational agreements). Itbegins far ahead (e.g. six month) before the flights for managing the scarceresources which are the runways capacities in relation with the airlinedemands. In a second step, Air Navigation Services Providers (ANSPs) areinvolved for adjusting the first version of the operational agreement to theATC resources and finding the best solutions in the district airspace. The thirdstep is a refinement and update process for managing the disruptions beingable to modify the second version of the operational agreement. Operationalplan is a continuous process which leads to the deliverance of contracts ofobjectives at each flight before its departure from the airport block. The

Advanced Concepts


operational plan aims at increasing the decision making process in theelaboration of contracts of objectives by a better transparency and datasharing.

5. Decentralized ATM organization

European airspace is naturally inhomogeneous in terms of demand. Differentmodes of operations will prevail in different parts of the European airspace.Modes will correspond to different qualities of service in relation with thetraffic density and the available technical and human resources. This conceptwas proposed first by ACARE (2004) and is totally consistent with the SHIFTvision. It is a new way for considering the ATM over European airspace, and itrequires studies for determining the modes of operations and the interfacerules between them. In this way operation modes are not only based ongeographical characteristics and traffic topology but is also time depending. Toassume efficiency (to optimize resources) the operation mode varies duringthe day according to traffic densities and resources availability.

In order to handle this diversity better, it would obviously be beneficial toadapt operational paradigms at local level. This would result in a EuropeanATM system where the provision of navigation services would be based onvarying operational modes dependent on location. This local tacticaladaptation would seem more consistent and advantageous overall than theaircraft-by-aircraft adaptation which is sometimes mentioned. We can imaginea system where traffic is globally coordinated by a strategic body which issupported by a set of districts managed independently by tactical bodies. I nthis system, the tactical districts are both the partners, when strategy isdecided (schedule preparation), and the local managers of the strategy’simplementation. The current resource management system views resourcesas fixed and handles traffic accordingly. Conversely, one could equally naivelyconsider the request to be sacred and subsequently identify the resources toaccede to it. A compromise between these two situations needs to be found,based on a policy decision between free access and performance.

The resource management of ATC in relation with the traffic demandsrequires the district-based airspace would be adaptive. In this case, it isresponsibility of the local ANSP in charge of the district to determine the bestbalance between its local resources, the traffic demands, and the chosenairspace solutions (airways, flight levels or waypoints). Then, the ATM globalorganization is decentralized and the ANSPs have the authority and theresponsibility of their choices for a greater efficiency.



6. Target windows

At the European level, the size of the airspace and the traffic diversity conductto share the Air Navigation Services responsibility between different actors.To assume responsibility, there need to have a significant autonomy in termof organization, as presented in the Decentralized ATM organization. Toensure a global coherence between actors concerned by the airside,intermediate objectives have to be negotiated. The target windows definemilestones marking out traffic progress. These intermediate objectivesassigned to Air actors have the following functions:

They constrain traffic progress in term of boundaries. At the strategiclevel, they permit to define or refine the airspace at the local levelaccording to expected traffic density and ATC capability. These opendynamicity in Airspace domain.

They create a strong link between the planning phase and ATCoperations increasing robustness of the whole system. The nature of thelink has to preserve ATC initiative and windows are to be calculatedaccording to the balance between constraints, disruptions and costs.Robustness comes also from the fact that local means are not directly theaim of the negotiation but only objectives on interfaces are discussed.

The collaborative planning on objectives permits to take into accounttechnical and economical diversity of actors and give guaranties. TargetWindows create add-values to technical and economical organizations.

The Target Windows create a common language between all the operatorsinvolved, and between the planning and the operations. Target Windows area tool that defines efficiency objectives for the operators, and provide amonitoring tool at tactical and strategic levels, enabling them to deal withdisruptions as soon as possible and with a clear view of the situation. Ratherthan precise 4D points, they are expressed in terms of intervals of adaptedwidth. Their size and localization reflect constraints faced by downstreamcomponents, such as punctuality at destination, runway capacity, or congesteden-route area. The room for adaptation left to operations ensures resilienceto disruptions. Operational divergence from this planning frame is stillpossible, and triggers a specific decision process at strategic level calledrenegotiation.

7. Dual airspace

The traffic complexity in the "core area" requires defining a specific mode ofoperation by separating the various types of flight, i.e. climbs, descents andover flights, in which the traffic is segregated into flow-based traffic anddistrict-based traffic. The aim here is to propose an original air traffic

Advanced Concepts


management system which will enable to cope with the peaks in demandexpected in the future. It is reasonable to suppose that the increase indemand will result in an increase in en-route traffic over the core area, whichis already congested to the point where it gives rise to numerous regulations.The aim is to relieve pressure on the main traffic axes forming part of thecore area's interlinked network by setting in place highways independent ofthat network. The highways will span the continent, and they will be reservedfor steady aircraft in level flight. Traffic management on highways will be flow-based, with closure conflicts but no convergence conflicts. In the core area,the theory is that there will be a limited number of highways along the maineast-west and north-south axes. Highway intersections generate no routesconvergence since they are managed through different level allocations. Thehighway constitutes a privileged airspace with tremendous potential forinnovation. The aircraft which make use of it are stable at a given level; theyall move in a predictable manner, in the same direction and the same way. Anaircraft flight path, then, is no longer three-dimensional but mono-dimensional. The possible benefits of this situation are threefold andcomplementary:

simplified traffic, making it possible to assimilate a larger number of aircraft;

simplified ATC, through the use of a limited system of elementaryinstructions: change in speed and change of route (digital airspace);

simplified displays, replacing the map backgrounds with synoptic tables.

The district-based traffic will be specific to local traffic in order to cope withthe local constraints of traffic and airspace. Regional airspace should turn to itsadvantage the isolation of a significant proportion of cruising traffic:

direct decrease in the volume of traffic;

increase in the reliability of predictions;

functional specialization of districts (climb or descent);



increase in airspace availability.

The paradigm SHIFT Operational Concept Document defines in detail thedifferent concepts raised in the first phase of the project [13].

8. Research agenda

These concepts requires further theoretical, operational and sustainabilityanalysis. Some questions like responsibility issues of this highwayimplementation, or more quantitative assessment of some parameters liketraffic throughput, should be added. The questions addressed by these fivemain concepts have been listed in the form of a Research Agenda.Investigation topics are identified and structured into interrelated stand-aloneexperimental research topics, with a multidisciplinary approach.

As shown in the figure above, the Contract of Objective could bedecomposed in different themes of research concerning the operationalcontinuity, acceptability from ground side and on board.

If we talk about Operational Plan, the problematic will be divided into threemain aspects: the different actors involved in its elaboration, the process onitself and the infrastructure required to support it. The Target Windowsconcept will address the questions of disruptions and the operational decisionmaking to deal with. To demonstrate the concept of Decentralized Design,the modelling of the interactions between Airspace design and TargetWindows could be imagined. Some quantitative assessments could be done

Advanced Concepts


on strategic traffic. The Dual Airspace will introduce the Highway and all thequestions linked with (frame, localization...) and consequently the cohabitationenvisaged between districts and highway.

9. Conclusions

Considering the amount of new and innovative concepts proposed, we dobelieve that the one year brainstorming session has been the right approachto innovate and generate new ideas. The grouping of expert having differentbackground brought fruitful input in the concept definition. Now, thechallenge for us is to demonstrating the relevance and the validity of conceptsin the frame of safety, capacity and efficiency issues linked to the growth in airtraffic in Europe after 2020. The validation work should focus on:

1. the observation of the impact of the suggested Contract of Service for anANSP;

2. and the measure of the influence of a suggested Highway in a classicsectorised airspace, in terms of human acceptability, workload, safetycriteria, capacity and efficiency.

10. References

[1] ACARE (2004). ATM team: High level ATM concept for the year 2020,V0.2. Advisory Council for Aeronautics Research in Europe

[2] ACARE (2002). Strategic Research Agenda. Volume 2. October 2002.Advisory Council for Aeronautics Research in Europe

[3] CDM (2003). Airport Collaborative Decision making ApplicationApplications: Operational Concept Document. EUROCONTROL -EATMP Reference n° 030408-01, Edition n°1, February 2003.

[4] Gottlinger, W. & Fakhoury, F. Airport CDM at Barcelona Airport:Collaborative Decision-Making at Barcelona Airport, Note No 03/02,EUROCONTROL, 2002.

[5] Florent, J-P. & Delain, O. Airport CDM at Zaventem Airport: CollaborativeDecision-Making, Improving Airport Operations through CDM, Rev 1,EUROCONTROL, 2002.

[6] Cormier, H. (2004) « Pour un système mondial de gestion du traficaérien. » (XIeme Conférence Mondiale de la Navigation Aérienne )Aviation Civile n°324 Juin 2004.

[7] EUROCONTROL (2004). Eurocontrol Operational Concept Document(OCD) , Volume 1 (the vision). European Air Traffic ManagementProgramme. Edition 2.1, 12 January 2004



[8] Garot, J.,M,. & Ky, P. (2003). The future air transport system in Europe:vision and perspective. AIAA/ICAS International Air & Space Symposiumand Exposition, Dayton, Ohio,USA, 2003.

[9] Gate to Gate (2004). Report on the Gate to Gate Integrated OperationalConcept (consolidated description). Validation of a European Gate to GateOperational Concept for 2005-2010, contract EC n° GRD2/2000/30120.29 April 2004.

[10] Gawinowski, G., Nobel, J., Grau, J.Y., Dohy, D., Guichard, L., Nicolaon, J.P.& Duong, V. (2003). Operational Concepts for SuperSector. FifthUSA/Europe seminar on ATM Research and Development. 23-27 June2003, Budapest, Hungary.

[11] Graham, B. (2004). Focused Vision Concept Input to Cooperative ATM.EUROCONTROL ATM R&D Symposium 2004, 14 June 2004, Stockholm,Sweden

[12] Grau, J.,Y., Gawinowski, G., Guichard, L., Guibert, S, Nobel, J., Dohy, D.,& Belhacene, K. (2004). SuperSector Experimental Results: Proof ofConcept Assessment. 23rd Digital Avionics System Conference, October24-28, 2004, Salt Lake City, Utah, USA.

[13] Guichard, L., Guibert, S., Hering, H., Nobel, J., Dohy, D., Grau, J-Y.,Belahcène, K. 2004. Paradigm SHIFT Operational Concept Document,EEC Note No. 01/2005.

[14] University Concept Team (2003). Airspace and Airports Concepts. Reportof the University Concept Team, Airspace Capacity Program, AmesResearch Center, NASA, Moffett Federal Airfield, CA 9403, USA.

FAME, Future Air Traffic Management Performance Enhancement


FAME

FUTURE AIR TRAFFIC MANAGEMENT PERFORMANCE ENHANCEMENT

EC - 6FP STREP proposal 2004

Gilles Gawinovsky, Marc Brochard

EUROCONTROL Experimental Centre, Brétigny, France

Considering the ACARE Strategic Research Agenda (SRA 1) [1] stating that itbecomes mandatory to propose an innovative Air Traffic Management (ATM)solution coping to the challenge of traffic demand (2020 horizon) [2], andimproving the efficiency and safety level of the European Air TransportSystem. An innovative new ATM concept was elaborated: Future Air TrafficManagement Performance Enhancement (FAME). It was proposed toenhance the ATM system via the adoption of a time based systemarchitecture and processes providing synchronisation between the ATM chaincomponents. This concept would lead to a more efficient network and flowmanagement between high-density airports. The key features to exploit thisconcept are a dynamic layered planning providing the basis of 4D contractsfor aircraft trajectories, a new concept of automated support tool and systemoperator task sharing for Air Traffic Control Working Positions. This 4Dcontract will thus be negotiated and ‘signed’ between aircraft operators or theaircraft itself and the Air Traffic Management system and other relevantactors. In these 4D contracts, 4D trajectories will be assigned to the aircraftto ensure an acceptable level of traffic complexity and density. Changes to thecontract will be managed either automatically by the issuance of a newcontract, or by specific actions that will ensure minimal impact on the trafficsituation.

1. Rational for the innovative FAME concept

ICAO forecasts a growth in world air travel of 5% per annum until 2005.Based on recent experience in Europe, this appears likely to be a conservativeestimate for this part of the world. Variations, regional as well in type of traffic,do occur; e.g. commuter traffic growing faster than long-haul. It is, therefore,not unreasonable to expect that air traffic in Europe may almost triple in the2002/2020 timeframe, as stated in the Vision 2020 [2] and the ACAREStrategic Research Agenda [1].

Advanced Concepts


It becomes clear that at the 2020 time horizon, revolutionary measures are anecessity to respond to the group of personalities Vision 2020 objectives [1].The future ATM system will be increasingly automated whilst the human willretain a significant, but different, role. In this future system, traffic trajectoriesare to be predicted in accordance with the level of system capabilities;defined new roles and new responsibilities of human and machine, groundand air elements addressing uncertainties and changes. Such a challengingobjective can only be met through a combination of actions that will betargeted at removing the inefficiency in today’s way of managing air trafficconsidering the Air Transport System as a whole. Some of the foreseensolutions are more of an institutional or regulatory nature. As such, thesesolutions need to be made possible through infrastructure, exploitation andtechnological enablers that are part of this project.

Revolutionary steps will be made possible by the ability to move towards amore predictive environment, enabling progressively higher degrees ofautomation. In such a system, adaptive and reactive mechanisms shouldcomplement the predictive parts in order to manage the uncertainty. A keychallenge will be to investigate the human role versus the automation of thesystem and its evolution, understanding human capabilities, ensuringcontinued improvement of capacity and safety, whilst maintaining jobsatisfaction.

In consequence, a time-based ATM architecture is envisioned, based on acontinuum of predictive-adaptive-reactive layers in order to move to asynchronised ATM system, and also on a move from a conflict-based to atime-based ATM approach (network and flow control). The major enablerswill be 4D Predictive capabilities, automated support tools for Air TrafficControl Working Positions, highly dynamic and synchronised layers, workingmethods based on layered planning approach, dynamic and flexible airspacemanagement (tubular airspace) and 4D ATC trajectory and contract ofservices. This 4D contract will be negotiated and ‘signed’ between aircraftoperators or the aircraft itself and the Air Traffic Management system andother relevant actors. In these 4D contracts, 4D trajectories will be assignedto the aircraft to ensure an acceptable level of traffic complexity and density.Changes to the contract will be managed either automatically by the issuanceof a new contract, or by specific actions that will ensure minimal impact onthe traffic situation. This 4D contract would enable:

Improving predictability: up to the departure of the aircraft, a contract isnegotiated and signed between the aircraft operator and theinfrastructure service providers. This contract should be constrainedenough to provide a real filter to the situation, but also flexible enough toincorporate the intrinsic uncertainty of the system. For instance, the



contract may stipulate mandatory requested times of arrival on specificway-points, but be flexible on the rest of the trajectory.

Reducing complexity: there is a need to identify how contracts can beupdated aiming at minimising traffic complexity. This is envisioned to bethe role of traffic managers, whose main task will be to organise trafficflows in order to minimise the traffic complexity, and thus the problemsdensity. This task will require an extremely high flexibility and adaptability,and is likely to be executed by humans rather than machines, since theywill occur up to 10-20 minutes in advance.

Reducing Uncertainty: even with higher predictability and reducedcomplexity which should lead to reducing the conflicts, if it illusory to saythat there would be no more conflict. Few conflicts will remain and willneed corrective actions. These will be solved on the basis that, forinstance, 10 minutes in advance, the degree of precision of the orderswhich are given to the aircraft is sufficiently high so as to be perfectlyreliable. Thus, the operational concept will be based either on groundcontrol, which will assign precise 4-Dimensions 10 minutes trajectories tothe aircraft, or on airborne separation functions. It is likely that approachesto airports will be operated by autonomous separation and meteringfunctions, such as station keeping.

In consequence, the future ATM system would be able to manage efficientlythe overall complexity with higher predictability (towards a more predictive4D system), while adding flexibility and closed loop information (layeredplanning functions, aircraft self-separation and station keeping). The systemwill move from a paradigm of reactive traffic control (acting at the local sectorentity level) to one oriented traffic management and able to manage thenetwork and flows. It should enable the overall objectives of moving moreaircrafts through the available infrastructure.

Also, the air transport industry is focusing on technology, while the otheractors are focusing on areas as short-term operational procedures andairspace improvements. Consequently, there is only little investigation towardthe future 2020 horizon addressing the problem in a holistic perspective,using the powerful capabilities of airspace features in a highly dynamic andsynchronised layered mechanism and 4D predictive environments. Both theEuropean Commission and EUROCONTROL have conducted researchfocusing on one part of the technology, procedures or airspace triangle, butnot so much in a holistic and innovative way. Airspace is very often notconsidered while main efforts are placed upon technology and/or procedures.

Is the airspace structure the key to pave the way to a more capacitive andsafety system? It is recognised that one of the key issue is the ability tomanage the network, and therefore the ability to manage the flow (according

Advanced Concepts


to the traffic pattern). In consequence, the European airspace still needs to bere-considered and re-designed at a European global level and not locally only.In such an approach, highway route structure should be considered withpowerful features as flight level allocation scheme and offset or parallelmechanisms. This promising approach should be able to provide extra-capacity and to transform this extra-capacity into safety, i.e. more aircraftdiluted in a much more capacitive airspace, while opening the floor toautomation development.

4D trajectory capabilities have been considered in different ways from full-adherence to estimated time of arrival, but the challenge to have a rigid 4Dcontract (prediction) while conserving flexibility and room of manoeuvre toadjust the situation coping with unpredictable events and uncertainty has yetto be investigated. In fact, integrating the adaptive mechanism in the predictiveone (i.e. the relative 3D navigation inside the absolute 4D navigation) appearsdifficult to mix. A more 4D predictive system is based on the accuracy andcapabilities of aircraft trajectory prediction model. From this assumption, it willbe possible to envisage a higher degree of automation. The challenge facedfor the past 20 years have been to define the right place and interfacebetween the automation and human roles, task and responsibilities. Thisquestion of the human role in a more accurate (if not perfect) informationsystem is still pending. Therefore, this issue in transferring the cockpitautomation philosophy and know-how into the ATM domain needs to beaddressed.

2. FAME concept

We can conclude that the major limitation of the present ATM system is theloss of effectiveness due to the weak interactivity between the two maincontributors: the strategical ATFM (imprecise long term planning) and thetactical radar controller (very accurate local reactive mode). This couldexplain the “capacity wall” due to the inability to manage the network andflows. We need to create a continuum between the predictive and reactiveparts in proposing medium-term anticipative layers with automated supporttools which will enable the synchronisation of all the layers (predictive,anticipative, and reactive) and therefore the management of the network.This concept will synchronize ATM operations and result in efficiency gains.The challenge consists in the successful transition from the today’sasynchronous ATM to a tomorrow’s synchronous ATM one which isexpected to provide extra-capacity. Thus, the FAME concept takes us froman essentially non-synchronous and conflict-based system constraint by thetactical controller cognitive capabilities to a better synchronized (using allavailable capacity in a more effective manner) and highly automated



(providing tighter, more accurate control loops) system providing flexible andadaptive mechanisms to adjust the predictive part. The Operational Conceptwill encompass CNS/ATM concepts and capabilities, structured around thefollowing major enablers:

Airspace design and management;

4D Contract Collaborative Layered Planning;

Automated support tool.

2.a. Airspace design and management

Rather than operating under a “free flight” with unstructured route network,it appears that the requirements for managing highly structured, synchronized,and scheduled traffic flow would need a highly constraint route networkwhich should remain still flexible and dynamic. Traffic assignment onto theairspace will have to be more directive in order to solve strategically asignificant proportion of problems; when a better use of the entire airspacewill be guaranteed by using predefined geographical routes. This couldtypically be done by segregating the airspace according to different modusoperandi. This development of airspace structure would enable dynamic andflexible capacity management through re-routing, to be executed at thestrategic, pre-tactical and tactical level.

We proposed to structure the airspace in such a way to manage efficientlythe network and flow management while taking into account the optimizationof the en-route and terminal approach interface. Route network topologiesare a mix of point-to-point high-density highway route network and switchtrunk route network. The first one is dedicated to specific dense city pairswith none other route interaction and the second represents a grid coveringthe European core area. These highway/trunk have a 3D tube structure(tubular airspace) based on flow-oriented adapted to traffic pattern, parallelroute to support offset capabilities, flight level allocation scheme (FLAS) andlarge volume sector. It will support the strategic de-confliction with no maincrossing flow and application of FLAS, and the tactical organisation withparallel routes.

The operating philosophy is to use a type of exploitation for the high-densitycity-pairs in pushing traffic through the tubes to optimize the scarce largeairport capacity resource, and to use an other type of exploitation to managethe switch between different trunk flows. The dynamic and flexible airspacemanagement will be supported through:

Traffic assignment onto the airspace will have to be more directive inorder to solve a significant proportion of problems at a strategic level;

Multiple route option (using predefined geographical routes);

Advanced Concepts


Pro-active and real-time coordination between all the planning layers;

It is proposed to define generic airspace design and management solutionscompliant with a time-based architecture which presents high potentiality forextra-capacity and safety gains. It means that the airspace infrastructureconstraints would:

Increase drastically the theoretical prediction values and thereforerepresent the main support to improve the predictive component of thisconcept;

Provide an increased theoretical capacity;

Create safety buffer aiming at increasing the safety gains.

From this approach, it is proposed to design highways between city-pairs andswitch trunk route network in Europe. Next step would be:

To map and find the best fit of the generic route network with theEuropean traffic pattern;

To optimise the en-route network to the terminal approach area; To connect to the secondary route network (small airport);

To assess the capacity, safety, efficiency and robustness of the system.

2.b. 4D Contract Collaborative Layered Planning

The main hypothesis of the proposed time-based architecture is to have aseamless 4D predictive and adaptive mechanisms. There are two mains issues.The first one is to develop algorithms to provide a 4D conflict-free trafficdistribution. The second one is concerning the network buffer design, e.g.room of manoeuvre to have some flexibility inside the rigidity of theprediction in order to take into account the uncertainty and unpredictableevent. This system relies on the fact that in the case of contract objectivesdeviation, the flexible and adaptive layers will be able to come back to theinitial contract of service or to renegotiate contract of services withoutdisturbing the network and reach its Request Time of Arrival. The buffercharacteristics are related to the network and flow synchronisationmanagement, i.e. the way the adaptative strategy will be conducted. Wepropose a flexible framework for managing highly structured, synchronised,and scheduled traffic based on layered planning functions in order to identifydifferent levels of air traffic management:

Strategical conflict-free prediction concerning the strategical ATFM partand aiming at assigning predictive 4-Dimensions trajectories to aircraft.The strategic traffic planning ensures quantitative management of theaircraft flow (i.e. traffic distribution) on a time base of 2 hours before take-off.



Managing the network and flows concerning the tactical ATFM and aimingat performing tactical and qualitative flow management with ananticipating time threshold between 20 and 30 minutes. This is applied bya traffic balancing mechanism.

Synchronising the traffic inside the flows concerning the anticipative layerand aiming at ensuring aircraft spacing (En-route ASAS) according to thecontract of service.

Aircraft separation: It concerns the remaining conflicts which are out ofthe nominal situations

This raises several challenges such as:

Defining the efficient layered planning represents the fundamental basis forthe definition of the working methods. It needs to define carefully thedecoupling task level, task, roles and responsibilities as well as informationexchange, planning action and time ranging.

Assignment of the predictive part to the automation while keeping theadaptive and flexible high-added value to the human. And therefore, themain issue of the human role in this loop and the predictive-adaptiveinterfaces.

Efficient automated support tools for Air Traffic Control Working Positionin accordance with the ATM predictability level.

2.c. Automated support tool

A more detailed analysis of the working procedures of an air traffic controlleror a team of air traffic controllers working a sector compared to otherdomains shows an astonishing picture. For example, working procedures in amodern glass cockpit were analysed, and significant differences were found:the Pilot Flying, who is responsible for the primary flight control, concentrateson decision making and monitoring, while tasks like planning andcommunication are carried out by the Pilot Non Flying. On the ground, theExecutive Controller is not only monitoring the safety relevant processes, butin addition concentrates on communications as well as on short term planningand decision making. Along with this situation, there is the widely acceptedstatement that the Executive Controller is in a critical load situation and isseen as a bottleneck for the needed capacity increase. The typical answer tothis situation from the Air Traffic Management (ATM) community is tosupport the Controller with additional tools in order to reduce his/herworkload. This approach has not lead to a significant change at the Controllerworking position, mainly due to the following reasons:

The implementation of additional tools is very often generating additionalcommunication overhead and additional need for interaction. In today‘sworking procedures there is no share of work between operator and

Advanced Concepts


system. Every support tool has to be fed with input data by the controllerand has to be monitored, which generates additional workload. This effectis contradictory to the needed relief at the Controller working position.

The different tools that are foreseen for implementation will create atbest a few percent capacity increases each. This possible effect is notvalidated for most of those tools, especially where a combinedimplementation of different tools is foreseen.

The basis for most of the Controller tools is the prediction of the 4Dflight path, which is either updated through Controller input or – asplanned in the future – updated through data link from the aircraft’s flightmanagement system. Prediction accuracy of 4D Profiles is limited asController clearances as well as the time a clearance is given are notpredictable. Even accurate Flight Management System 4D Profiles, ifavailable from A/C, will not lead to the necessary accuracy unless thecontroller’s intent is integrated into the prediction process.

The introduction of automated support tools for the Controller workingposition, where routine tasks are transferred to a system component, is partof the FAME approach with COSTA (COntroller Support ThoughAutomation). In contrast to many current ATM R&D approaches, COSTA isbased on a re-organisation of ATC tasks, ATC roles and an ATC 4Dtrajectory tailored to the needs of applications with enhanced accuracy inflight path prediction – the “ATC Intent Trajectory”. This will lead to thedevelopment of new procedures concerning the man-machine interactionand the design of an HMI supporting the new system-operator taskdistribution. By these means, a change from the well known paradigm ofdecision support in ATM towards a new paradigm of automation will evolvetogether with a corresponding evolution of operator roles. The finalimplementation of the concept will enhance the level of automation to level 3to 5 [3].

3. Conclusion

This proposal did not pass the evaluation thresholds for eligibility and thus hasbeen rejected by EC. Even if the topic addressed was felt as very importantfor the future of ATM, it did not shown going too far beyond current state ofthe art. Internal review within INO concluded that we were probably tooambitious in our proposal with a not mature enough concept. It was felt thatthe proposal was done too early.

Thus, for future proposal, we should first refine any proposed concept, checkcarefully its maturity and then, stress more on the objectives of such aninnovative ATM concept and the added values which could be gained.



4. References

[1] ACARE (2002). Strategic Research Agenda. Volume 2. October 2002.Advisory Council for Aeronautics Research in Europe.

[2] EUROCONTROL (2004). Eurocontrol Operational Concept Document(OCD), Volume 1 (the vision). European Air Traffic ManagementProgramme. Edition 2.1, 12 January 2004.

[3] Sheridan, T.B.: Telerobotics, Automation and Human Supervisory Control.Cambridge MA, USA: MIT Press 1992.

Tube Advanced Lane Control


TUBE ADVANCED LANE CONTROL

Jean-Pierre Florent, Marc Brochard

EUROCONTROL Experimental Centre, Brétigny, France

1. Introduction

When the Military close (use) their five airspaces over the Maastricht area,the delay in the area is more than 1 hour per flight and automatically thetraffic is concentrated in air traffic lanes. One question raises immediately: canwe organize these traffic lanes, decrease the average delay, and expedite moreflights?

Tube Advanced Lane Control (TALC) is proposed for one attempt trying toanswer this question. It describes a new concept of highway aiming atimproving the efficiency of the Air Transport maintaining the safety in acrowded airspace. This concept develops a new airspace design and a newoperational mode in a reserved area: the tube.

A new distribution of tasks and roles is proposed between the main actors ofthe air transport: i.e. Airport, Airline, Pilot, ATCo and their relations withautomated systems. The project assesses also the impact of a collaborativedecision making applied to the air fully derived from the CDM airportconcept: new data, new processes for better adapting resources to the trafficdemand.

2. Tube Automated Line concept

Several studies point out limit inherent to the human being to manage moreand more traffic. With an increasing demand, the human being will becomemore and more a limiting factor in that traffic growth. Amongst variousresearch axes to design tools, methods and/or infrastructures that wouldrelease this limitation, several ideas are investigating the possibility of giving ordelegating more and more tasks and responsibilities to computer systems. Inthat trend, the computer would replace the human being and would executetasks managing the traffic.

The Tube Advanced Line Concept (TALC) is one attempt to designcomputer assisted control concept keeping the controller as the ultimatedecision maker, and associated to advanced airspace organisation (the tube orfreeway). The objective is to increase capacity and punctuality in a context ofsustainable air transport development

Advanced Concepts


Capacity would be increased by reviewing flight management in specificairspace areas. The trend is to consider flows (or a group of flights) as oneobject instead of managing individual flights. This flow management wouldimpact air traffic control rules and would support automation. This new wayof controlling flights would also require a new airspace design: the tube orfreeway where aircrafts will flight on dedicated axes cruising at dedicated flightlevel. Entry/exit and expedite windows would link the advanced airspacestructure with standard sectorised airspace which should allow a seamlesstransition.

In terms of punctuality, we shall consider that airlines wish: first to have in-time flight meaning flight to be as close as possible to the estimated in blocktime at destination; second dynamic management between trajectoryprediction and needs meaning prioritization, punctuality where their partners(Airport, Handlers and ATM) give advantage to designated flights in case ofdelays. The airports (ADEP and ADES) have their constraints (capacity,airport slot, sequences…) and wish coordinate the flights trajectory to thereal time gate to gate conditions. The TALC concept assumes that advancednegotiation phase will take place for each flight amongst all Air Transportactors to get the best flight profile and scheduling to fit at best to the variousconcerns of the various actors involved for each flight.

In this area, the freedom of the actors (controllers & pilots / airlines) and theprivate performance of an actor are taken into account but come later on theperformance of the entire network: more capacity, more safety, morepunctuality, and less pollution.

The natural places for such control are between area pairs where the trafficdemand is getting higher. When we have a look ahead of ten years or more,the number of saturated areas will increase. The TALC concept does notpropose a big bang or to abandon the sectorised control. The sectorisedcontrol and the tube control share the airspace, work together to expeditethe traffic demand. The resources of the ATC are traditionally the airspacevolumes (sector) and the ATCo (or team of controllers) and associatedoperational working methods. The TALC concept impacts these classicalresources by adding a new airspace volume: the tube; and new operationalworking method associated to the new airspace volume (including new set ofresponsibilities amongst new set of actors). These shall ensure an optimalprocessing of aircrafts operating in this new airspace volume, and also asmooth transition and completion with classical sectorised airspace areas.

This concept shall not be seen as a big bang, but rather as a help for theactual sectorised control in high density flight areas. Also, in any cases, anyuser will have the freedom to operate in a tube or in a classical sectorised



area. In order to optimise the operation in tube and to cope with automation,only certified aircraft will be allowed to operate in this new airspace area.

3. The TUBE airspace model: TAM

The Tube Airspace Modelling shall consider the building of new airspace areaaiming at managing high density traffic between 2 specific areas (be it city pair,TMA pair or navigation points pair). The tube is supposed to support conflictfree flow with flight anticipated management process keeping safety as thefirst objective. The tube becomes one of the smallest controlling entities,replacing the sectors in a specific area.

The design of the tube airspace starts with the specification of an idealtrajectory taking into account the airlines preferences. Airlines preferences aredefined according to their flight policy: aircraft type, economical flight or just intime flight defining the Flight Economical Weight (FEW). The FEW can varyaccording to the crew management, the number of passengers, the numberof the connected flights (Hubs); and, in real time, it depends of the scheduledeviations or of the accumulated delays. According to the local needs andrequirements (distance, level, speed, airlines intentions…) the actorsdetermine a subset of trajectories which will become the standard flow withina tube. In order to get the optimum in term of flight operation, a standardflow will be a given flight level, and flight level changes will be minimised asmuch as possible. The best case would be not to have any flight level changesexcept cases where safety becomes critical. According to the density of trafficwhich would make use of the tube, there might be several flows or flightlevels usable by the aircrafts in this tube.

As tube is designed to get more capacity with the maximum of safety, thetube concepts proposes to push the automation as a mean to optimize trafficproceeding. This automation would tend to reduce the human workloadwhich is, as stated earlier, part of the capacity barrier. From the ATCo pointof view, we know that the majority of his/her work is to avoid conflict. Tubewould be designed to provide a conflict free area by offering organised trafficwhich would by construction and organisation avoid any crossing (lateraland/or vertical). All flights flying over a tube flow will follow the samedirection at a cruising flight level. Therefore, any flow will be unidirectionalwith no crossing areas. From the aircraft (aircrew) point of view, flightseparation will be automated in order to have train of aircrafts auto-separated. Of course, this would need specific operation rules to support thistraffic organisation and to ensure high safety level (see Tube OperatingModel).

The design of the tube shall also consider the traffic density and the trafficforecast in the years to come, in order to evaluate the entry and exit point of

Advanced Concepts


the tube. One can imagine seeing a tube linking city pairs, but, past studies, forinstance TOSCA, have shown the poor interest and poor added values insuch cases. Nevertheless, tube can be a link between 2 high density terminalareas and/or a mean to increase capacity in high density area by developing akind of freeway with the objective of segregating the traffic. In any case, atube will be materialised by at least one entry point and one exit point. Thesepoints will be the connection between tube airspace and sectorised airspace.According to the traffic complexity, its density, and its nature (be it city pair orareas pair), a tube could be more sophisticated and providing several entryand/or exit points.

As a tube shall be connected in any cases to a sectorised area, the entry andexit phases of a tube shall be very carefully processed in order to ensure asmooth and safe transition between the 2 kinds of airspaces. Within a tube,flights are going to follow standard flows at predefined flight levels. Any time,a new aircraft will have to enter to the tube, there will be a transition phaseto accommodate: the transfer of the flight between the sectorised and tubeairspace, which means also a transfer of responsibility and operating rules andthe sequencing of the flight to enable to meet the entry conditions agreed forthis flight (see Tube Operating Model). This operation needs both accurateplanning and anticipation but also given airspace area to enable the entrysequence. This airspace area is the tube entry lane. We could imagine havingat least on entry lane per entry point, but in some sophisticated tube, wecould imagine having several entry lanes for one entry point.

In order to consider the diversity of the situations in terms of trafficcomplexity and the external constraints like peak hours, the tube structurehas to be adaptable, flexible according to the traffic forecast, be it plannedtraffic over the time or real time traffic facing to unplanned event (badmeteorological flying conditions for example).Therefore, as summary, a tube will be composed of:

One or several entry and exit points which will make the connectionbetween the tube and the classical airspace (sectors);

Unidirectional and non crossing flows ensuring conflict free navigation andsupporting automated navigation;

Entry and exit lanes ensuring the smooth entry to or exit from the flows.

4. The TUBE Operational Model: TOM

The tube concept has 2 main objectives: increase capacity and punctuality.Capacity would be increased by the new airspace structure itself, but only ifalso supported by dedicated and optimised operation rules: the flowmanagement procedures. Punctuality in opposition of capacity would be



increased with new operation rules but above all, by optimised planning andsequencing. This means that punctuality objective starts before the tube entryitself, when capacity would start at that time. On top of these 2 elements ofthe new procedures required for the tube operation, the automation willbring new transfer of responsibility which will impact the overall tubeoperation processes.

Therefore, the Tube Operational Model covers:

The tube flight profile definition: CDM-A concept; The tube flow management;

The tube operation automation.

5. Collaborative Decision Making concept: layeredplanning approach for Tube Flight Profile definition

As stressed before, airlines wish above all to have in-time flight meaning flightto be as close as possible to the estimated in block time at destination. Thisinvolves many actors and an efficient sharing of accurate flight informationamongst all actors involved in the flight. This process is derived from theCDM airport concept applied for the air segment. Bearing in mind that tubedoes not cover the entirety of the flight (the entire flight path beingcomposed of classical sectorised airspace and tube areas), the response tothis challenge is to take part of the overall negotiation about the flightscheduling from block to block, and from this negotiation, to derive the besttube flight profile: this be the entry, exit times and the best flight profile(especially in terms of flight level) for a given flight, with the responsibility forthe tube manager, to stick to this commitment. For this negotiation phase,and for each individual flight, the following constraints and requirements shallbe considered with all the actors:

With the airline:

Flight Economic Weight (FEW) status giving the prioritisationbetween flights

Preferences (level, speed, punctuality) Environment & Pollution

With the airports:

Estimated take-off & Estimated landing time, SID, STAR

Coordination Tube / Airport (Planning of estimated entry/exittime in the tube)

With the ATC:

ATFM slot

Advanced Concepts


Tube and flow maximum capacity and punctuality (level andspeed)

Coordination Tube / Sector (Planning of estimated entry/exittime in the tube)

Anticipated Clearances (ACLs): next manoeuvre to perform(vs. controller’s input order)

Automatic Conflict Detection and Resolution (vs. tactical orderof controllers)

Alarms (disruption, trajectory deviation…)

This phase can be seen like an extension of the Collaborative DecisionMaking system (CDM) applied to the air segment. This assumes that a strongcollaboration will be put in place between ground and air segment and actors.The collaborative process aims at anticipating the demand as much aspossible and at refining the tube flight profile and the entry/exit conditionsthrough an iterative process: the layered planning approach. This processmight be as follow:

Airline submits the flight plan 3 hours before the Estimated Off BlockTime (EOBT). From that time, a process of coordination between thetube and the airports (departure and arrival) starts. The first tube entryand exit conditions for this flight are defined with an accuracy of 15’. Thetube capacity might be adapted to cope with demand. When the trafficorganizer detects a lack of capacity is detected, supplementary flows areactivated.

around 20 minutes before the start-up of the flight, the accurate TargetTake-Off Time (TTOT) taking into account all the departure constraints ismade known to the tube manager. This information is crucial for startingthe coordination between the tube and the sectors, and also to reviewthe tube entry-exit conditions for this flight with an accuracy of 5’.

20’ before the flight enters in the tube or at lest 20’ before the ActualTake-Off Time (ATOT) when the tube entry is very close to thedeparture airport. The tube entry-exit conditions for this flight arereviewed with an accuracy of 1’30’ and the 4D flight profile is computed.



T.Milest

TUBE

Entry Exit ALDT

FPL- 3 hours

EOBT TSDT

T-DPI- 20 min

- 20 min

TTET TTXT

ELDT- 0 min

Tube Data Quality

-15’ -5’ -1’30’’

CPTM

CTPM

APTM CPTMAPTM CPTM

Tube

Flow

Trajectory

Air -Side : Tube Entry Time & Landing Time

AOBT

Tube - Capacity & Punctuality <> Flows NumberCoordination Tube – Airports <> Tube Entry Window 15’

Flow <> Coordination Tube – SectorTube Entry Slot 5’

Trajectory <> User’s NeedTube Entry Time 1’30’’

Layered Planning Approach – Air Side.

6. The tube flow management

Tube flow management includes both tactical and strategic aspects. From atactical point of view, there is a need to manage and adapt the flow capacityaccording to the demand. As a tube shall be flexible, flow capacity will not bethe same at any time. Fluctuation will occur in order to adapt the flowcapacity to the demand, and to minimise as much as possible the impact ontothe sectorised airspace. For a strategic point of view, a tube must react to anydisturbance that could occur in real time. Disturbance might be due tounplanned event in a tube (meteo, delay), or changes occurring outside thetube but impacting the tube flight profile (like for example, capacity reductionat destination airport which will need tube flight profile reconciliation withnew airport capacity).

As soon as the tube flight profile is agreed, the sole responsibility of the tubemanager, after safety, is to guarantee the respect of the entry and exitconditions which are time and flight level. For nominal cases, its tube flightprofile being negotiated and agreed before hand, each flight will be segregatedin terms of nominal speed and flight level that shall ensure the respect of theexit time and a conflict free navigation. In case of delay due to the tube flowmanagement, the tube manager will have to operate the flight and/or the flowso that the overall tube flight profile is respected, and thus on time exiting theflight from the tube; or to negotiate a tube flight profile update with all actorsinvolved for this flight.

Advanced Concepts


The following constraints shall be considered for an optimal tube flowmanagement: Tube-Flow:

Maximum Tube and Flow Capacity and punctuality (level andspeed);

Flying time policy: Just in time (i.e. Absorbing of the delay) orEconomical flight according to users request;

Characteristics of Entry & Exit points: number, length, position,entry-exit sequencing

Ground and land sides:

arrival and departure airport slot,

airport capacity and traffic regulation vs. capacity

Air side:

Tube Flights profile at Entry and Exit

Prioritization between airline flights (FEW)

Airport Slot Reconciliation (between flight delay and airportslot)

Conflict Detection & Resolution

Passage from one flow to an other flow (tube with differentoffsets, levels & speeds)

Exit in a seamless way from tube area to sectorised area(coordination, disruption)

Function: en-route, sequencing, overtaking, crossing with routeoffset (lateral or vertical)

7. Tube Operation automation

As stressed several time, automation and new task and responsibilities willcharacterise the tube concept:

Automation will be allocated to the tube flight profile negotiation like 4Dtrajectory definition (including tube entry and exit conditions), flightsequencing to enable conflict free navigation, and users needmanagement. These tasks will be delegated to a new actor: the computerwhich will process the tube flight profile definition;

4D flight profile monitoring and keeping will be delegated to the airborneside: i.e. the pilot. As tube shall ensure conflict free and unidirectional



navigation with no crossing, airborne separation and collision avoidancewill be delegated to the airborne side.

Tube and flow management will be performed by the tube controller.He/she will be responsible for respecting the tube flight profile and, moreespecially, the disruption management. Coordination with sectorisedairspace will be also under the responsibility of the tube controller.

Each of these actors will have to coordinate theirs tasks in a collaborativemanner.

• Delegation to Computer

4D TrajectoryTrajectory Conflict Less Anticipated ClearancesFlight SequencingUsers Needs Management

Controller

Pilot

Computer

Info

rmat

ion

Decision

Control

• Delegation to Airborne

Airborne Separation Airborne Collision Avoidance

• New Tasks for the Controller

Tube ManagementFlows ManagementSector – Tube CoordinationDisruption Management

Each Angle

. Needs Information

. Makes Control

. Takes Decision

8. Research Plan

Today, Tube Advanced Lane Control is only a proposal for a new concept.Not really new as it is the continuation of several works undergone frommany years. The initial activity would be to refine the concept by focussing ona strong definition and analysis of the new responsibilities and roles betweenthe various actors involved (airport, airlines, controller, pilot and their relationwith automated systems), having in mind the treatment of non nominal cases(safety). Business cases analysis including economical cost benefits analysis,environmental impact, and sustainable development analysis shall be alsoperformed in order to assess the overall benefits of the TALC concept.

This done, the second part of the research plan would be to focus on theimplementation of the concept in terms of services and data required tosupport the processes, and then to develop various model of the new system(macro and micro models) to support different level of simulation. This phaseshould consider:

Advanced Concepts


The enrichment of shared information and the definition of new data byreviewing and increasing the quality and the accuracy of the existing data.This task shall start from similar concept already existing like CDM airportand may lead to define new data not existing today. It will be alsomandatory to get a common agreement of information state ofoperations (data: sharing, updating...).

Develop new data to define airline preferences and economical weight,airport constraints and arrival status, tube capacity, constraints, entry/exitconditions and tube flight profile.

The specification of new mechanisms of exchanging information betweenground and air side, and for supporting layered approach like FlightUpdate Message, Departure Planning Information, Airport SlotReconciliation or Tube Coordination with Sectors, Airports and Airlines;

The operating tools to be used in the tube increasing performances andenabling automation. Existing tolls shall be used as much as possible likeDMAN and AMAN to be linked with tube tools, safety nets like MTCD& MTCR (Medium Term Conflict Detection / Resolution), or FMS,ACAS, ASAS. Other specific tools may be needed according to newproblematic caused by tube and flow control.

The layered approach for defining tube flight profile like Flight Prioritisationat Airport, Pre-departure Sequence, Flight Prioritisation in a tube,Adapting rule or criteria for sequencing (first call, first served or any othersequencing policy), Auto-Regulation in crisis situation at airport, DynamicATFM slot based on Target Take-Off Time or Slot Swapping and Shiftingbetween all flights.

9. Conclusion

Given the level of automation and the difficulties of implementing this TubeAdvanced Lane Control, the operational deployment might be medium orlong term. For a medium term deployment, only part of the concept could beconsidered with probably a low level of automation and a localimplementation. Even if the objectives of this concept are to increase the en-route capacity and the punctuality, the innovative part lies more in theprocess of sharing information between all partners involved for a given flight,this including ground (airport and airlines) and air sides. Tools and data exitalready today, the matter is to share this information and to makecommon/collaborative decision. Flows management, tube and automation areonly means to support the overall objective of punctuality. Safety isnevertheless the major constraint and shall not be under estimated.



The main challenge of the Tube Advanced Lane control is to reconcile theobjectives of all partners and to steer towards a common goal: to provideand to get the best service.

10. References

[1] EUROCONTROL 1999, Testing Operational Scenarios for Concepts inATM, TOSCA Phase II Final Report, Assessment of the TMA to TMA hand-over concept (WP3), August 9, 1999..

[2] EUROCONTROL (2002), Florent J.P , EEC Report No: 371. ImprovingAirport Operations Through CDM (March 2002)

[3] ACARE (2002). Strategic Research Agenda. Volume 2. October 2002.Advisory Council for Aeronautics Research in Europe

[4] CDM (2003). Airport Collaborative Decision making ApplicationApplications: Operational Concept Document. EUROCONTROL -EATMP Reference n° 030408-01, Edition n°1, February 2003.

[5] EUROCONTROL (2003), Zaventem Airport, Collaborative DecisionMaking - Improving Airport Operations, EEC Report No. 371, Mar. 2003,

[6] EUROCONTROL (2003), Florent J.P , Collaborative Decision Making -Malpensa - WP 2, 18 Nov. 2003,

[7] EUROCONTROL (2004). Eurocontrol Operational Concept Document(OCD) , Volume 1 (the vision). European Air Traffic ManagementProgramme. Edition 2.1, 12 January 2004

[8] ACARE (2004). ATM team: High level ATM concept for the year 2020,V0.2. Advisory Council for Aeronautics Research in Europe

[9] EUROCONTROL (2004). EATM Performance Enhancement Activities,Edition 2004

[10] EUROCONTROL (2004), Ehrmanntraut, Rudiger, The Potential of SpeedControl, Jul. 2004

[11] EUROCONTROL (2004), Ehrmanntraut, Rudiger, Measures of AircraftConflicts in Europe with a Simulation Model, Jul. 2004

Full Automation of ATM for High Complexity Airspace


FULL AUTOMATION OF ATM FOR

HIGH-COMPLEXITY AIRSPACE

Rüdiger Ehrmanntraut

EUROCONTROL Experimental Center, Brétigny, France.

The objective of this concept is to significantly increase overall ATM systemperformance with the full automation of the controller function, i.e. anincrease of capacity beyond levels of human operability with guaranteed andpredictable safety levels and reduced costs. The work focuses on airspace ofhigh complexity where the shortfalls of the current air traffic control willappear first. The additional required capacity will be provided by improvingthe medium term planning function by better traffic organisation and byreducing the complexity of traffic with the help of algorithms. The approach isto get a better understanding of capacity bottlenecks by analysing trafficcomplexity and medium-term conflict, and to research possibilities for themedium-term traffic organisation that would overcome these bottlenecks.Other performance parameters like safety, environment and economy areconsequently evaluated.

In the year 2005 several studies have been carried out that resulted in anumber of publications, from which one has been awarded as best paper.

Work on complexity has resulted in an update of the statistics about ECACwide conflicts as would be produced by the flight plans from theCFMU [R. Ehrmanntraut, R. Christien, 2004, Measures of Conflicts in Europewith a Simulation Model, EUROCONTROL Experimental Centre, inproceedings of the 23rd Digital Avionics Conference DASC, Oct. 24-282004, Salt Lake City, USA]. One entire day of traffic was simulated with theComplexity Light Analyser (COLA). The specificity is the use of correctedflight plans, i.e. flights are updated when the effective trajectories are deviatingmore than a specified tolerance buffer from the initial flight plan, using newdata from CFMU. The results are that only between 9 and 18% of conflictsare horizontal encounters, depending on the flight level. In addition most ofthe conflict geometries are precise parallel same and parallel oppositedirections. A mathematical discussion interprets the empirical results underthe angle of uncertainty for the prediction of conflicts. It finds that manyconflicts do not suffer from low predictability because of their encountergeometries.

Advanced Concepts


climb-descent28%

descent-descent17%

climb-climb14%

cruise-descent14%

cruise-climb18%

cruise-cruise9%

Figure 1: Conflict Attitudes from FL 60 to 180.

cruise-cruise18%

cruise-climb30%

cruise-descent21%

climb-climb7%

descent-descent9%

climb-descent15%

Figure 2: Attitudes above FL 180.



Further, the conflict density has been studied for the core European airspace[R. Ehrmanntraut, 2004, Complexity Of Speed Resolutions - Conflict Density,in proceedings of the 3rd INO Workshop, Dec. 2004, EUROCONTROLExperimental Centre]. An extended review on literature for complexity andconflict resolution finds that the conflict density has not been defined, nor wasit used beyond the simple counting of predicted conflicts per controlledsector. Therefore it is useful to investigate the spatial and temporaldistribution of predicted conflicts. The study proposes different categories ofconflict density by relating the predicted conflicts to sectors, routes, waypointand other volumes created by the conflict geometry. Then several simulatedtraffic scenarios covering the core European area and using traffic growth upto the year 2025 are analysed, using the Reorganised ATC MathematicalSimulator (RAMS).

Figure 3: Number of overlapping protection buffers at closest point ofapproach of conflicting aircraft, defined as conflict-CPA density.

Advanced Concepts


Figure 4: Number of conflicts on routes leading to waypoints NTM andARCKY, defined as conflict-on-route density.

The graphical visualisation of conflict densities seems to be very useful tool.The results show that problems mainly occur at fixed areas in the proximityof some waypoints of crossing routes and on some routes without crossingtraffic but high traffic densities. The analysis of conflict and resolution densitiesgives strong arguments to microscopic airspace modifications around hotspotsinstead or in addition to global airspace design. It could also give argument todynamic route network changes that adapt to specific traffic and conflictpatterns during the day. In addition it was felt that conflict density is a helpfultool for the analysis of model simulations in that it can visualise theproportion of resolved and unresolved conflicts and herewith give a visualsupport for the improvement of the resolution algorithm.

Further work has focussed on the resolution of predicted conflicts with theuse of speed control [R. Ehrmanntraut, 2004, The Potential Of SpeedControl, in proceedings of the 23rd DASC, Salt Lake City, Utah, USA]. It wasfound that speed clearances are not used for en-route air traffic control,especially not for conflict resolution. Therefore the study elaborates thepotential that speed control could have to resolve predicted conflicts in adense environment. Again the setup of the RAMS simulator was used for theEuropean core area and with extrapolated traffic samples to the year 2025.



The simulator had to be adapted for the resolution by speed, and the aircraftperformance model had to be improved. The simulation setup would allowfor speed variances up to +-15% within the performance envelope of theaircraft, and would only move one aircraft by conflict resolution. Thecontroller in this model looks 15 minutes ahead beyond the sector boundaryand starts to apply speed manoeuvres about 14 minutes before the predictedconflict.

-

2 000

4 000

6 000

8 000

10 000

12 000

14 000

B B B B

100 150 200 300

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

Conflicts Unresolved % Resolved Unresolved / Flights

-

2 000

4 000

6 000

8 000

10 000

12 000

14 000

B B B B

100 150 200 300

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

Conflicts Unresolved % Resolved Unresolved / Flights

Figure 5: Number of conflicts, unresolved conflicts, resolution rates andunresolved conflicts per flights for speed control.

The result is that speed control is very effective and can resolve 70% for themost demanding environment by applying very low speed adjustments withonly 4% reduction and 5.3% increase on average. Further analysis reconfirmsthat many conflicts have parallel opposite and parallel same directionencounter angles where the speed manoeuvre is less efficient, but could becomplemented with other manoeuvres like parallel offset.

An interesting discussion is the use of speed control in an automation system.The results are promising and the advantage of speed, especially with its smallspeed adjustments, would allow the automation system to issue speedclearances without the sector controllers recognizing the system actions andherewith lead to a dramatic reduction of workload.

The study will continue on both axes, i.e. the analysis of complexities and theelaboration of resolution manoeuvres for the planning time horizon.

Analysis of the Impact of Small Aircraft on ATM in Europe


ANALYSIS OF THE IMPACT OF SMALL AIRCRAFT

ON ATM IN EUROPE

Daniel Rohacs(1,2), Marc Brochard(2)

(1) Budapest University of Technology and Economics, Hungary.

(2) EUROCONTROL Experimental Centre, Brétigny, France.

General aviation, and air traffic of regional airports (or the airfields which donot offer a conventional radio-location approach) need a new, cheap, andsecure control system. This demands totally new, innovative ideas, and theusage of the newest technological achievements that could be applied to theentire system in an integrated fashion. From the commercial aviation practice,the system philosophy could be used as a base, than it should be modifiedand applied to general aviation needs.

First, the current system, and the market differences between commercialaviation and general aviation are reviewed. After the description of thecurrent and future state of aviation, the market information shows us thegood timing of the research. Secondly, a possible usage of a new aircraft isexplained. Such an airplane should be able to be piloted by everybody(without any special or extra knowledge), just as easy, as we driving our careach day. Therefore that requires a high level of automation, and airborneseparation assurance, especially for the control of mall, or/and emergencycases. Naturally that also means the transfer of separation responsibility fromground to air. Finally, the airport requirements for small aircraft usage havebeen described.

1. Introduction

Why do we need a new system? The answer seems to be simple. [1] Today’sair traffic volume is projected to be double by 2020 [2]. As the existingsystem is already reaching its limits, for tomorrow’s capacity it will not be ableto meet future needs in several areas: airport, airplane, environmentalconsideration, security, and safety, etc [2].

The regional flight in new democratic countries, in new members of EU mustdevelop rapidly by the increasing economy fact [3]. However the regionalairports in those regions are not equipped with modern radio-locationsystems for controlling the air traffic.

Advanced Concepts


On the other hand, today, there are [3],[4] 300 000 private and small aircraftpilots in Europe. They fly more then 60 000 small aircraft. Annual market ofgeneral aviation is 5,5 Milliard EURs only in Europe. With accordance to ourinvestigation this market will increase even with greater ratio thenconventional air traffic.

2. Future state of aviation

All that means, we will be obliged to use a much more efficient system in theaim of saving the world from the large economic costs of flight delays andcancellations.

It is said, that today is the perfect time for the industry to develop the airsystem of the future, because just in a few years ahead air traffic will bereaching record levels, so till then we have a little time to prepare ourselvesfor the work ahead of us [1]. And this time is not too far off. When looking ata 30-year worldwide trend, by the Gulf War and the Asian financial crisis thegrowth of air traffic slowed down only temporary, exactly like after theSeptember 11th terrorist attacks. So this drop in air traffic expected to beshort lived and traffic growth seems to be returned at record levels.

After we have seen the world annual traffic growth, probably it would be alsointeresting to understand the reason of this increase.

0

200400

600

800

1000

12001400

1970

1975

1980

1985

1990

1995

2000

2005

2100

year

incr

eain

g (1

00 %

rel

ated

to

year

196

0)

GDPcargo trafficpassanger traffic

Figure 1- Economical development effects on traffic needs.

In practice, the needs in transportation are rising rapidly [6]. Each percent ofincrease in GDP generates a growth in: people mobility 2 and in cargo traffic2.5 per cent respectively (see also Figure 1). During last decade (1990 –1999) the length of highways was increased by 25% [8]. Especially in well-developed country, like USA [9], the highway system and number of cars



nearly stabilized. However passenger-miles of aviation are increased 14 timesduring last 40 years [6].

The high-speed railway transport and intelligent highway system concepts arenot certainly the best solutions for future transport [6],[10]. They can increasethe road and railway transport capacity, but they have a much greater landdemand for realizing the transportation and even the influence and load onthe environment seems to be more important than in case of airtransportation [10]. So at near future, we could develop a new airtransportation system for wide public use. Such type of new air transportationsystem could change completely the future world and could improve thequality of people’s life.

In any case, clearly, the technology is available to establish a safety, economicaland environment friendly new air transportation system.

Figure 2 - Travelling habits : current, and future state [11].

In this future where air transportation will be reaching high level limits in trafficgrowing, the freedom of peoples, their ability/willing of travelling where andwhen they want, will become the more and more important. Aviation has toensure it, so probably the travelling habits will change [11] (see figure 2). Thevision is to expand the current aviation state along all dimensions. For thiswork, probably the most important is to ensure a more flexible airtransportation. Instead of the currently used hub and spoke system, travellingon demand, and/or point-to-point travelling will take over the lead. Thisfuture state can only be reached, if aviation uses radically new, innovativeideas, while breaking down the currently existing limits.

Advanced Concepts


3. Some market information

The reason and the problem of the increasing fact of air transportation areimportant, but we should also examine the future aircraft deliveries to have awider approach. The outlook [5] covers global demand for commercialaircraft and business operation throughout the world.

Figure 3 - Future aircraft delivers (business aircraft).

In case of business aircraft, a total of 23.000 jets will be delivered over thenext 20 years to fractional and tradition operators (see the figure 3). Amongthem, around 8000 will be microjets. (By the definition of the source [5],microjet has a maximum take-off weight between 5000 and 10000 lb, likeCitation Mustang, Honda Jet, Avocet, Safire, etc.) The economic slowdownand related uncertainty that affected the world’s economy during 2002 and2003 made the manufactures to reduce the production. As a result of thisdecision, deliveries are expected to be higher in the second part of thisdecade. As we can see, this microjet demand level will continuously be veryhigh for the next decade [5].

The next figure (4) shows, that the delivery units (for microjets) willapproximately be 37%, but the delivery value will only be 5%. With otherwords that means a high level usage of small and cheap aircraft. Naturally thatwill be a very good market not-only for aircraft providers, but also for thewhole air transportation.



Figure 4 - Delivery units, and delivery values for future aircraft type.

Even with the non-examination of other problems (like situation-awarenesscaused by a possible same cruising altitude of microjets and other commercialaircraft), the lack of a new ATM is already visible.

4. Research focus

In the introduction we have already seen the problem caused by theincreased traffic volume in the future. However the impact of small aircraft onthat saturated air transportation system seems to be a more complex,thought provoking problem.

To ensure the future needs, several domains have to be enhanced, such asdeveloping airport infrastructure, introducing new policies (noise, emission,environment protection, etc), use of new materials, fuels, and design process,creating a responsible air traffic system, etc. In the aim of having a wide viewof the current system, each problem has to be understood clearly. Howeverit is not possible to focus on all of the difficulties, after a time we decided tostudy the air traffic management and air traffic control area.

The development of a new system, where air traffic controllers, pilots, andother users are able to have more precise information about the aircraftposition, and motion is a highly increasing demand. Thanks to innovation andthe newest technological achievements, such a system could have anincreased choice in the integration of several new domains.

Advanced Concepts


The NASA has initiated already a similar project [7], [11] called as smallaircraft transportation system (SATS). The NASA program is focused on thenew aircraft design, airports development and economical foundation of theproject.

We think, the personal aircraft has to be designed for using by commonpeople. So, we need principally new aircraft, which could be piloted byeverybody, without any special or extra knowledge and abilities. With otherwords that means people who have not been well trained, even so theyarrive to fly the airplane, simply as easy as they driving their car [4],[6]. Suchaircraft will be used very widely. Therefore we have to develop radically newsystems, and instruments to be able to accomplish tasks as pilot workloadmonitoring, simplified control [4], etc. One of the most interesting problemsis the automatic control system (especially if we widen our ”definition” forsmall aircraft by UAVs). That means for example a possible control for mallcases (the system could review pilot’s decisions, and even take away thecontrol, while realizing an emergency autoland).

Naturally this high level of automation requires advanced airborne systemapplications. Advanced situation awareness can be ensured by airborneseparation assurance, which transfers the responsibility of separation fromground to air (meanwhile changing the role and responsibilities of ATCOs).The cockpit instrumentation and on board systems must be characterized byhuman-aided automation that will provide intuitive, easy to follow flight pathguidance superimposed on a depiction of the outside world. Softwareenabled flight controls and flight planning will increase single-crew operationalsafety and mission reliability to two-crew levels. To help this goal a keyelement could be an enhanced (artificial/synthetic) vision, and highway-in-the-sky 4D guidance. (note – artificial vision could also be an equipment toreduce weather impacts on small aircraft)

If the small aircraft is flying in non-radar airspace and around small non-towered airports, flexible or optimal usage of airspace will become important.That probably requires a usage of under-utilized airspace, and newprocedures. Nevertheless that can create increased capacity at nearly anylanding site in Europe (Figure 5 [7]). To solve the possible problem ofcommunication, small aircraft should optimize the ground and airborne dataprocessing systems, with the creation of a global interoperablecommunication, navigation, and surveillance system. In this field, newtechnologies like airborne internet communication standards and protocolsfor client-server communications, multi-channel secure communication, orspace- and ground-based sensing could be used.



Figure 5 - New era of small aircraft.

All this means that the new cost-effective, customer-orientated system shouldgive precise traffic information to all possible users (ATCOs and airspace aswell). With other words, we have to ensure peoples to able to travel whereand when they want, in a more safe and affordable (travelling cost should beless from passenger view) way. To reach this goal, the development of newsystems and subsystems is indispensable.

5. Objectives

One of the objectives of the research is to obtain a wide knowledge of thefuture state of small aircraft transportation. A general problem of this study isto identify, and then to understand the differences between providers’ ideaand users’ requirements. If these differences are already named, we couldadopt them to the current air transportation system and/or air trafficcontrol/procedure.

The main objective is to forecast/analyze the impact of the future smallaircraft on ATM, and airspace construction. After this study a new conceptwill be needed (probably), that will pose high requirement on automationcapabilities. An increasing aircraft autonomy should be supported by thetransfer of certain procedures and responsibilities from ground-based toairborne separation assistance, that has two main benefits: changes the roleand responsibilities of controllers, and improved flight situation awareness.

Advanced Concepts


6. The new aircraft type

To ensure the airplane needs described already, we can find several smallaircraft families on the market. The market analysis shows that a 4 – 9 seatedaircraft is required. They can be equipped by a jet, or propellered engines. Ofcourse, the noise of the engines has to be reduced, in case of close aircraftlanding and/or take-off to the city center.

Figure 6 - One of the possible new small aircraft that can influence the ATM.

For aircraft requirements we could use the ones described in the NASASATS documents, which are the followings [6], [7] :

the target operational environment will be in class C, D, E, G, F. Maybethe class A, B as well, with Required Navigational Performance (RNP).

speed at least 200 to 300 knots

cruise altitudes of 6.000 to 25.000 feet range : 800 – 1.200 nautical miles

graphical flight-path operating systems

new engines

automatic control systems

etc.

To fulfil these requirements, we must not forget the fact, that principally forthe last 40 years, the small aircraft applying the latest results of the sciencesand technology have not been developed. So we need absolutely newdesigned aircraft. The main objectives of the small aircraft development arecharacterized by the followings:

developed aerodynamics – even use of revolutionary concepts,

principally new designed engines – with reduced fuel consumptions, noiseand air pollution, (possible diesel engine and jet solutions),

excellent performance – for good and safety piloting,



excellent ride control – with application of active and adaptive controlmethods, good technical life – with reduced fatigue damages, use ofdamage tolerance design philosophy, etc.

Using these constraints for the future small aircraft requirements, the currentlyavailable airplanes could be the followings (see figure 4.1.) : Eclipse 500,Honda Jet, Lancair Columbia 300, Cessna Mustang, Avocet ProJet, AdamAircraft A500, Adam Aircraft A700, Safire Jet, Diamond D-Jet, etc. Each ofthem has a range between 980-1250 nm, a maximum altitude around41000ft, and a 340-380 kt cruising speed. Most of them use the latesttechnological achievements (such as General Aviation Propulsion program[12]).

Note. Made by the different geographical size, probably there will be adifference between the US and the EU used range, and so the cruisingaltitude as well.

7. Small aircraft in current airspace structure in Europe

To have an idea whether an interoperability is required with the commercialair traffic (caused by the same cruising altitude), an airspace structure studywas needed. With other words, that means the study of the cruising altitudes.

The cruising altitude depends on several values, where the most importantare the followings (figure 7) :

temperature, pressure,

humidity,

take-off weight,

flying distance,

the optimum of speed and consumption,

wind, etc.

We also realized that airplanes in general try to get the more and morealtitude [14], in the aim of having lower fuel consumption at high flight levels.Even so it is not worth climbing up like a rocket, then “falling down”, airlinecompanies try to have an optimum between climb and cruise, which effectscruising altitude diminution. For short-haul flight, cruise can vary from 40% upto 60% [14].

After the preliminary calculation and average flying distance study, we havefounded out, that the cruising altitude will not reach more than FL 180 in caseof 500 nm maximum range.

Advanced Concepts


Figure 7 - Cruising altitude reaching time.

So we can conclude that the small aircraft for our goal will not have aninfluence on commercial air transportation (see figure 8) while the cruisingpart of the flight (small airplanes will not fly over, and/or pass by the samealtitude).

While we were studying the airspace structure, we have met anotherproblem, which is the lack of a harmonized airspace system [15] in Europe.Without a future organization across Europe, small aircraft users will not beable to use a single continuum of airspace. Their freedom will also be limited,by the national boundaries, which not allows an optimum altitude at all thetime.

For wide propagation of the project, strategic steps towards a simplifiedairspace organization have to be ensured. A good example is the EuropeanSingle Sky project, where initial steps towards a unique European system havebeen already taken.

8. Possible airports for small aircraft in Europe

One of the reasons of using small aircraft in the future is to reduce the door-to-door time. Therefore a new net of small airports is needed.

These airports [4], [6] should have radically reduced volume, where forexample, the runway length could be 1500 m at maximum. They also should



be built more close to city centers, which allows the users to reduce theirtravelling time to the airfield. Naturally noise restrictions must be stricter.

The typical airport for small aircraft could have a structure shown in figure 9.Generally, in two direction, if it possible two – two parallel runways could bebuilt. The major elements of such a "smart" airport of the future are thefollowings [2], [7] :

New airports more close to the city center,

Precision approaches to all landing areas,

Flight Information Services (FIS), broadcast by terrestrial or satellite systems,

Traffic Information Services (TIS),

Destination Information Services (DIS)

Near-all-weather operations at non-towered airports without radarcoverage,

Safety services

Figure 9 - Vision on the new small airport.

Before creating new airports, we should still consider the use of the existinginfrastructures. In EC we can find more or less 70 hub & spoke, 550 publicuse, and 2500 landing facilities. All that means a regional airport in each 150-200 km, a general aviation airfield in 50-60 km, and a landing facility in 30 km.

Advanced Concepts


9. Conclusion

From an ATM approach, an increasing traffic volume (especially in case ofsmall aircraft) demands enhancement in several areas. That could be flightcrew situation awareness improvement, and to change the role, andresponsibilities of air traffic controllers. These new technologies have to beapplied to the entire system intelligently and in an integrated fashion, in theaim of having a new, effective and a useful system. The future challenge of thisproblem must not be underestimated, because that will probably require newtechnological approaches, and innovative ideas.

10. References

[1] Airbus : Global Market Forecast 2003-2022, December 2003.http://www.airbus.com/pdf/media/gmf2003.pd

[2] Transportation Network Topologies Dr. Bruce J. Holmes, NASA; JohnScott, Icosystems, April 27, 2003.http://spacecom.grc.nasa.gov/icnsconf/docs/2004/01_plenary/PS-06-Holmes.pdf

[3] Regional Flight 2000, Hungary (project leaders: Rohács, J. and Gundlach,M., contractors: Hungarian and Bavarian Governments), reports I - III,BUTE - Budapest, Dornier - München, RHTW - Aachen, 1991-93.

[4] Rohacs, D. Diploma thesis : 2004 July, INSA de Lyon & BUTE : Nouveausysteme de controle automatique pour de petits avions.

[5] Rolls-Royce : Business Jet review and forecast, NBAA Las Vegas, October2004.http://www.rolls-royce.com/civil_aerospace/overview/market/outlook/downloads/busjet04.pdf

[6] Rohacs, J.: PATS, personal Air Transportation System, ICAS Congress,Toronto, Canada, CD-ROM, 2002, ICAS. 2002.7.7.4.1 -7. 7.4.11.

[7] Hahne, D.: The Small Aircraft Transportation System: A Potential Solution toFuture Transportation System, Workshop on Integrated CNSTechnologies for Advanced Future Air Transportation Systems Hostedby the Space Communications Program at NASA Glenn ResearchCenter, May 1st - May 3rd, 2001 Wyndham Hotel, Cleveland, Ohio,USA. http://spacecom.grc.nasa.gov/icnsconf/2001/agenda.shtml

[8] Transport infrastructure in Europe between 1990 and 1999.http://europa.eu.int/comm/eurostat/Public/datashop/print-product/EN?catalogue=Eurostat&product=7-09042002-EN-AP-EN&mode=download



[9] System Mileage Within the United States (Statute miles).http://www.bts.gov/btsprod/nts/Ch1_web/W1-1NEW.XLS

[10] Volker von Tein: Status and Trends in Commercial Transport Aircraft,Lecture on the ICAS'98 Conference, ref. number: ICAS-98-0.3,Melbourne, 1998.

[11] Dr. Bruce J. Holmes, NASA : Small Community Air Transportation,Oklahoma – CASI, November 14, 2003, OK_CASI_11-14-03.ppt

[12] General Aviation Propulsion Program (GAP)http://www.grc.nasa.gov/WWW/AST/GAP

[13] Eclipse 500 general information, and technical data.http://www.ainonline.com/Features/newbusinessaircraft02/nba_eclipse500.html

[14] Aircraft Performance Databasehttp://www.eurocontrol.fr/projects/bada/

[15] ATM Workshop, Airspace Strategy, Eurocontrol 20/11/2000.


EmergingTechnologies

Emerging Technologies


In this research the applicability ofemerging technologies to ATMproblems is being investigated.Emergent technologies can beinnovative in two ways: either bydramatically improving currentsystems/operations, these are oftenreferred to as d isrupt ivetechnologies, or by opening upcompletely new solutions whichprovide the required order ofmagnitude improvements for futureoperating conditions.

Most of the novel technologiesunder consideration are aboutvisualization and interaction with theATM system. The target of thisresearch is to improve the human-human and human-machineinterface performance. Theapproach is multi-disciplinary,combining computer science andhuman factors research.

Other studies address safety,security or cost efficiency issues,mainly (but not exclusively) from asoftware technology perspective.Such studies are motivated by therealization that the ATM system isto a large extend also a softwaresystem, and performance gains canthus be expected from softwareengineering innovations.

ATC Maastricht 2004Innovation Award

AIT adds an unnoticeable digitalsignature in form of a watermark tothe conventional air-groundanalogue voice communication, toincrease safety and security (INOActivity Report 2003, Watermark

Technology for VHF VoiceCommunication).

This Aircraft Identification Tagconcept was developed incollaboration with Graz Universityof Technology, Frequentis realized ademonstrator. At the annual ATCMaastricht event the editor Jane’shonored the AIT concept with their‘Innovation Award 2004’.

Strategic relevance

The ACARE Strategic ResearchAgenda strongly suggests thatinnovative research be undertakenon four topics, two of which areaddressed by this axis:

the concept of Tower-less ATM ;

innovative visualizationtechnologies.

The combination of both leads to afocus for the visualization andinteraction research on theapplicability of mixed realitytechnologies to the airport controltower. The fact that airports willsoon become the major bottleneckin the European ATM network is a



further motivation. One of themajor issues facing airports is their(often dramatical ly) reducedcapacity under low visibil ityconditions. Especially in thesec i rcumstances , v i sua l i za t iontechnology may hold its biggestpromise.

Visualization and interaction research

In 2004 a first series of studiesexploring 3D stereo technologywere concluded. Concrete resultsinclude:

improved interaction modes forpointing and selecting, “select-by-volume” ;

improved navigation modesreducing the risk ofdisorientation ;

effective visualization techniquesto exploit depth information,“drop-lines”.

In addition, this initial research hasled to fundamental methodologicalinsights:

The impact of any newvisual izat ion technique orinteraction mode can only bemeaningfully measured in thecontext of a specific control task.All recommendations from thisresearch have to be qualified withthe control task to which theyapply.

The impact of any newvisual izat ion technique orinteraction mode can only bemeaningfully measured throughcon t ro l l e d expe r imen t s ,comparing the performance of a

sample of users with a controlgroup, followed by statisticalanalysis.

Building on these insights, the focusof the visualization research in thenext 3-years period will be onAugmented Reality applications forthe airport control tower. Areas ofinvestigation are:

non-intrusive tracking for optimalregistration ;

visualization of airport objects, toimprove operations under low-visibility conditions andcollaboration betweencontrollers ;

task performance of towercontrollers.

In parallel, a framework is beingdeveloped to capture the results ofindividual multi-modal experiments.The fundamental objective of thisresearch is to understand theper formance of ind iv idua lmodal i t ies , to predict theperformance of their combinationsand to investigate their potential forincreased co-operative work. Withthis framework it will be possible totrace an optimal research strategythrough the huge design spaceoffered by the numerous noveltechnologies.

Finally, several 2D interfaces havebeen developed over the last 18months, which still need human-centered evaluation. Wheelieaugments the traditional syntheticradar display: by rolling the wheelon the mouse, the controller



highlights all flights on a referenceflight level. The expectation ofWheelie is a reduction in themental effort for the separationtask. Secondly, Vital tabularisestime-line representations of flightplan data. Vertical alignment ofwaypoints at the same flight levelindicates potential conflicts. Suchanalogue representations hold thepromise of faster reconstruction ofcomplex 4D situations. Severalpossible ways to extend Vital toprovide visualization support ineven more complicated situationsare already envisaged.

On-going collaborations

Central to the mode of operationof the innovative researchdepartment is the commitment toperform the invest igat ionswhenever possible in collaborationwith other research entities, therebystrengthening the scientific ATMresearch network in Europe andworldwide.

Since the beginning of thevisual ization and interactionresearch, collaboration with theNorrköping Visual ization andInteraction Studio of the LinköpingsUniversity (Sweden), headed byprofessor Anders Ynnerman, hasbeen fostered. Annual developmentiterations have resulted in acomprehensive test-bed for semi-immersive air traffic controlapplications.

The test-bed contains 3Dinterpretations of classical air trafficcontro l ob jects such as

representations for aircraft, flighttags, trajectories, or conflict zones.A d d i t i o n a l t e c h n o l o g i e sincorporated into the test-bedinclude sophisticated weathervisualization, terrain reconstruction,speech recognition as an interactionmode and, lately, positional audio. Inthe summer of 2004, two masterstudents from the computer scienceDepartment of L inkö pingsUniversity were detached to theEurocontrol Experimental Centerand developed a user-friendlygraphical preparation tool for flightscenarios.

The Flight Institute at DLR,Germany , fo l lows a verycomplementary agenda in its mixedreality research for ATM. Theirapproach is also based on amutually re-enforcing loop oftechnology-driven design andhuman-centered evaluation. Theirfocus is also the control tower andthey too have moved from initialVirtual Reality technologies toAugmented Reality applications.The potential for synergies is greatand will be further explored nextyear 2005.

Collaboration with Graz/Frequentis inthe context of AIT

The Aircraft Identification Tagproject continues the collaborationwith Prof. Gernot Kubin from the‘Signal Processing and SpeechCommunication Laboratory’ at GrazUniversity of Technology, Austria.Konrad Hofbauer, a PhD studentfrom this university, started his



research study in October. Hisstudy shall increase the capacity forthe AIT payload data withoutneglecting the other dependingparameters, inaudibi l i ty andreliability.

AIT’s digital watermark techniquecould open a way for adaptivechannel equalisation techniques.Such techniques are widely used fordigital telephone communication(GSM) to increase communicationquality. Beside this perceived speechquality another related researchtopic for a separate study could bepayload data security.

Participation in the 6th framework programme

In order to consolidate and extendthe network of collaborations, aproposal for the 6th FrameworkProgramme of the EuropeanCommission was submitted. Thisproposal, known as Virtual AirportControl Tower, aimed at exploringinnovative working practices andenhanced reality technologies forincreasing the throughput at airportcontrol towers, particularly underbad weather conditions. In additionto the established partnerships withthe Linköpings and Patras (Greece)universit ies , the consortiumconsisted of NLR (R&D / TheN e t h e r l a n d s ) , L F V(ANSP / Sweden), ThalesATM & TRT (Industry / France),DeepBlue (SME / Italy), andARMINES (University / France).

Although the evaluation report wascomplimentary about this proposal,

it did not make it in the (very short)shortlist. The fact that the budgetawarded to ATM within theAeronautics calls turned out to bedisappointingly small puts intoquestion whether resubmissionwould be worthwhile; although thiswas suggested by the evaluators.The innovative research departmentremains interested in participating inEuropean Commission projects, butwill not take the lead in any suchproposal in the near future. Inparallel, a watch is being maintainedon relevant European Commissionprojects in order to built on theirresults and learn from theirexperience.

Laboratory infrastructure

The visualization and interactionlaboratory is equipped with a Barcoworkbench, augmented with anIntersense ultra-sound trackingsystem and associated Steregraphicsshutter glasses. Powered by aSilicon Graphics Onyx2 with a singlegraphics pipe, this set-up isadequate for performing a range ofmixed reality experiments.

The lab includes classical 3Dinteraction devices such as a wandor a glove. A SensAble hapticdevice and ReachIn desktop display,combined with a 3D Labs graphicscard, provides force feedback along3 degrees of freedom and positionsensing along 6 degrees of freedom.

A camera-based eye trackingequipment from FaceLab (withpossibilities for video overlay) hasrecently been acquired. This



equipment provides non-intrusivetracking and auto-stereoscopicvision, with which a correctalignment of virtual objects on a realworld scene can be achieved. Thiswill form the basis for the plannedaugmented reality experiments.

Other research

In the second half of 2004, a highlyspeculative study has been kickedoff in which the potential impact of

the Open Source paradigm for theATM indus t r y , and fo rEUROCONTROL in particular, isbeing evaluated. The study hasidentified four hypotheses (in theareas of harmonization of systems,quality of software, business model,public service obligation) that willnow be explored.

Digital Signatures for Air-Ground Voice Communication


DIGITAL SIGNATURES FOR AIR-GROUND

VOICE COMMUNICATIONS

Konrad Hofbauer(1,2), Horst Hering(2)

(1) Graz University of Technology, Austria.


1. Introduction

The air-ground voice communications between air traffic controller andaircraft in a specific flight sector is done over an analogue VHF (Very HighFrequency) radio channel.

For identification, pilots have to start every message with their call-sign. Dueto various reasons there is a potential risk that the controller registers no orincorrect call-sign information (occurrence of so-called call-sign confusion).

A previous study demonstrated the feasibility of speech watermarking for theembedding of a digital aircraft identification tag into the voice communicationbetween pilot and controller. This system allows transmitting a short digitalmessage over the analogue radio communication link by adding an almostinaudible broadband signal to the voice signal.

2. AIT – Aircraft Identification Tag

Figure 2-1 shows the general outline of the AIT system as proposed in [1,2].Only the shaded elements represent AIT modules, whereas all other parts arealready existing aircraft equipment. Therefore especially the transceivers in theaircraft remain unchanged, which is an important issue in terms of systemcosts and certification.

Triggered by the PTT (push-to-talk) switch, the encoder embeds with arobust watermarking technique the data provided by the data acquisitionmodule into the analogue speech signal. This is transmitted via theconventional VHF transmitter to ground systems and surrounding aircraft.

These can receive and listen to the message without any special equipment. Ifthey are equipped with the watermark decoder, they can extract and displaythe data that is embedded into the signal. Integrated into the air traffic controlsystem, the aircraft that is currently transmitting could for example then beautomatically highlighted on the controller’s display.



Figure 2-1 – Voice communication link with embedded data [2].

2.a. Encoder

The watermark encoder is working in the digital domain and currently baseson direct sequence spread spectrum technology and frequency masking.

The first step in the encoder adds redundancy to the digital data by anECC (Error Control Code) scheme. This highly increases the reliability of thesystem and is necessary because of the distortions occurring in the VHFtransmission.

In the next step, this coded data is spread over the available frequencybandwidth by a well-defined pseudo-noise sequence. This watermark-signal isthen spectrally shaped with a LPC (Linear Predictive Coding) filter andembedded into the digitized speech signal, while exploiting the frequencymasking property of the human perception.

As a last step, the digital signal is converted back to analogue domain.

2.b. Decoder

After transmission and conversion to digital domain again, the decoder appliesa whitening filter on the incoming signal to compensate for the spectralshaping in the encoder. After the decoder’s synchronization to the datastream, the signal is de-spread and the watermark data extracted. With theredundancy included in the encoding stage, the decoder can correct errorsthat occurred during the radio transmission.



Figure 2-2 – AIT aircraft system architecture. The data is provided either fromthe flight management system, obtained by a radio receiver or permanently

set during system integration, e.g. the aircraft’s tail number.

2.c. Data Module

The purpose of the data extension module is to provide the payload data,e.g. SSR (Secondary Surveillance Radar) code, and PTT switch status to theAIT system. For simplified cockpit integration and certification, the data shouldbe acquired autonomously without connection to the aircraft’s internal databusses and without any user interaction required. Current research evaluatesthe feasibility of integrating two simple radio receivers into the data module,see Figure 2-2. One of them detects active VHF transmissions, which impliesthat the PTT switch is currently pressed. The second receiver continuouslymonitors the SSR identifier that is broadcasted by the aircraft’s transponder.Therefore it seems possible to integrate the AIT system into the connector ofthe pilot’s headset, without any further modification to the aircraft equipment.



3. Target Objectives

The defined aim is to improve the above system in regard to performanceand application range. The figure-of-merits of the system are mainly given bycapacity, audibility and reliability. These are mutually exclusive and limit eachother, i.e. a real implementation cannot perfectly fulfill all three criteria at thesame time.

3.a. Capacity

The capacity (or data rate) of the system describes the length of the digitalmessage in bits that can be communicated in a time interval of less than onesecond. Current research aims for a payload data rate of 100 bit/s, whichallows the embedding of both aircraft identification and aircraft position.

3.b. Reliability

The reliability of the system gives an estimate on the probability ofoccurrence of wrong data blocks. The ratio of erroneously received digitalmessages that are mislabeled as correctly identified aircraft identification tags,although the channel code actually fails to correct the errors, is intended tobe smaller than 10-4. This means that at most one out of 10000 transmittedmessages is erroneously shown as correct tag, although it is actually incorrect.

3.c. Perceptual Impairment

The watermarking process unavoidably introduces a small amount of noiseinto the speech signal. Where the watermark signal level exceeds thethreshold of audibility, the degree of annoyance for the voice communicationusers has to be kept minimal.

3.d. Security

Preventing impostors from faking the digital aircraft identification signature isan important security feature. This should be achieved with conventionalcryptographic security measures such as public key systems or synchronizedchaotic modulation and demodulation.

4. Near-term Research Focus

The ongoing study performs multidisciplinary research in communicationengineering, audio signal processing, psychoacoustics and adjacent fields. Inorder to achieve the above goals, the following research topics are studied indetail (work in progress).

4.a. Data Embedding Algorithms

The recent advancements in robust watermarking and data hiding algorithmsshow great potential to increase the possible amount of data that can beincluded in the voice message. Various algorithms are available which show



high robustness against channel noise and are tunable in terms of capacity,reliability and speech impairment.

4.b. Channel Models

In practice the data capacity is restricted by the poor quality of the band-limited transmission channel. But, the foundations for air-ground radiocommunications are exhaustively explored since the invention of radio morethan a hundred years ago. This contributes to find a good trade-off betweencapacity and reliability of the system.

4.c. Channel Equalisation

In the last decade, digital communications, as e.g. with xDSL4 or GSM 5

networks, achieved large progress through adaptive channel equalizationtechniques. Taking into account the parameters of the transmission channel,similar techniques may be applied for analogue radio as well.

5. Conclusion

Digital audio watermarking is a rather young and currently very active andevolving research area. Applications range from copyright enforcement andmeta-data systems for the entertainment industry up to hiddencommunications (steganography) in military environments. This study wantsboth, contributing additional knowledge to the watermarking community anddelivering an application for air traffic control. On various exhibitions andconferences, the presented application of watermarking to the VHF air-ground communication (AIT) is well received in the ATM world, due to itssimplicity, safety benefit and security increase.

6. References

[1] Martin Hagmüller, Horst Hering, Andreas Kröpfl, and Gernot Kubin.“Speech watermarking for air traffic control”. In Proceedings of theEURASIP 12th European Signal Processing Conference (EUSIPCO’04),Vienna (Austria), Sept. 2004.

[2] Horst Hering, Martin Hagmüller, and Gernot Kubin. “Safety and securityincrease for air traffic management through unnoticeable watermark aircraftidentification tag transmitted with the VHF voice communication”. InProceedings of the IEEE 22nd Digital Avionics Systems Conference(DASC 2003), Vol. 1, pp. 4.E.2-41-10, Oct. 2003.

4 xDSL : Digital Subscriber Line, includes ADSL (Assymetric DSL), SDSL (Symmetric DSL),HDSL (High data rate DSL), VDSL (Very High DSL).5 GSM : Global System for Mobile communications.

3D VR Air Traffic Management Project


3D VR AIR TRAFFIC MANAGEMENT PROJECT

Marcus Lange(1), Matthew Cooper(1), Anders Ynnerman(1), Vu Duong(2)

(1) NVIS, ITN, Campus Norrköping, Norrköping, Sweden.


1. Abstract

NVIS has been working with Eurocontrol's INO group for the last four yearsto explore the usability of 3D display and 'Virtual Reality' technologies in thesphere of Air Traffic Management and Control. NVIS' task has been toexplore the potential from the viewpoint of information visualization andinteraction and has produced four successive versions of an interactive, semi-immersive 3D visualization system for evaluation by INO.

Apart from the many internal changes made to the software to extend andimprove its function, common to any large software development which isunder constant redesign and extension, this year the work has focused onimprovements to many features of the system, such as weather and flightinformation display, but with a major effort to improve the interactionschemes provided with enhanced voice recognition and speech feedback andhand-based (data-glove) control systems being redesigned and re-implemented for improved interaction and control. In terms of enhancedfunctionality, the single largest addition is a new scheme for the detection offuture conflicts present in defined flight plans with real-time updates to permitthe controller to interactively update the flight plans to remove conflicts.

2. Major updates

A number of features have been enhanced and a number added during thisdevelopment phase. These individual features are each described below.

2.a. Flags

A range of information can be attached to the aircraft models representingeach flight in the controller’s sphere of influence. These data are representedusing flag objects.

Flag objects include both textual and graphical information connected to theairplanes by vertical lines, hence the term ‘flags’. The flags are sorted andordered in different ways so that they are visible at all times avoiding thevisual ‘clutter’ which is common in 3D displays of complex data.



A flag that is covered by another flag will be repositioned so that it is visible.Flags that are farther away are higher up. Stereo display resolves these flagobjects easily.

Flights can have different types including the name of the airplane, a graphicalor textual indication of the speed and the altitude. Individual flag objects canbe added or removed from the display, under user control, during runtime.

There are currently four different flag types:

Speed – indicating the speed as a horizontal bar.

Altitude – indicating the altitude as a horizontal bar.

Name of the flight – displaying the name of the flight as text

Destination airport – displaying the name of the destination airport.

Figure 2-1 – Flags displayed are sorted automatically as the camera tracks andaircraft positions are updated so that the user is always able to clearly read

them and discern which flag is connected to which aircraft.

2.b. More flight routes and airports

Up until now we have been focused on one particular airport, StockholmArlanda. Trajectories have been color coded differently for outgoing,incoming or unused trajectories.

When flight paths going from and to other airports were added to thesystem, the need for a fourth type arose, a color for a trajectory not



associated with the airport in focus. Also, when focusing on another airport,the color of the trajectories must change, to suit the new airport in focus.

For this purpose we have added an airports object. Each flight is defined witha ‘from’ airport and a ‘to’ airport. The focus can be shifted to either the ‘to’ or‘from’ airport as appropriate, making the camera change its centre of rotationto the selected airport. All trajectories and flights adapt their colors withrespect to the newly selected airport.

The flights database has also been substantially extended to include crosstraffic that does not directly involve our chosen airport of interest.

Some of the existing routes, which in previous versions of the system only leftArlanda and then disappeared from the controller’s sphere of influence after atime, have been extended so that they now have both a departure and adestination airport.

Figure 2-2 – Additional flights and airports allow the creation of morecomplex scenarios in the visualization system. The trajectories of flights

between airports not at the user’s current airport of focus are shown using aseparate colour (pink) so that they can be easily distinguished from the

incoming and outgoing flights (green and blue).

2.c. New speech client

In order to improve the speech recognition a new speech client has beendeveloped. The new client is, again, written in visual basic and uses therecently released Microsoft SDK 5.1.

It employs continuous recognition in place of the discrete recognitionprotocol that was employed in the previous version and provides a



substantially more reliable interface avoiding many of the sound environmentproblems and issues with microphone placement that afflicted the previousversion of the speech recognition system.

In short the recognition system has now reached a level of accuracy andprecision which permits its use for the majority of control functions with littlerecourse to pointer–based navigation and interaction.

Commands can easily be added to a configuration file.

Speech synthesis functionality has also been added to the client to enablevoice feedback from the ATM application.

2.d. Sound client: positional sound

The sound client connects to the ATC application as any other networkedclient and receives sound commands. A sound command consists of an ID,describing what kind of sound should be played, and the position andorientation of that sound in the 3D-world.

The client can either be run in text mode, doing nothing more than playingthe sounds, or in graphical mode where you can see from where the soundshould appear to be coming.

These sounds are presented to the controller using positional sound features,incorporating both stereo amplitude and phase modulation such that a verystrong sense of position can be observed when using a surround orheadphone system.

These positional cues provide a powerful attraction to the controller to guidetheir attention to the location of specific problems such as detected futureconflicts (see section 2.e and 2.f below).

2.e. Vortex wakes and Vortex trails

We have made use of data provided by Eurocontrol’s INO group to includevisualization of vortex wakes left by aircraft in the hope that such data canmake it possible for controllers to compress the flight approach paths to areasof concentration, such as are found at airports.

The approach taken in the data provided to us is to calculate the intensity ofthe disturbance at a series of points behind each flight described by eachaircraft of appropriate data orthogonal to the flight direction.

These data are affected by wind strength and direction as well as the flighttype. We interpret these volumes of data and isolate vortices where thestrength is above a user-defined intensity.

We then display the vortex as a tube following each flight. The data arestored and ordered with respect to time as the flight passes a certain position.



Figure 2-3 – Vortex wakes left behind flights are shown using colour codedlines. The line colours fade as the vortices grow weaker. Once their intensityhas decayed below a user-defined level they are no longer shown. The white

lines preceding the aircraft are ‘look-ahead’ indicators showing the flights’planned routes over the next few minutes.

In the current implementation, we have found that the data does not providesignificantly superior information over the standard, time/distance-basedseparation methods employed at present since the movement of the vortextrails is not visibly significant in our current display system. That is the vortextrails drawn appear as a trailed lines following each flight along its trajectoryand so do not provide significant additional information allowing thecontroller to compress the aircraft flow towards the point of convergence ofthe flights.

In future models, designed specifically for the airport approach scenario wemay be able to display this information with higher resolution, permitting theuse of this information in new ways which are, at present, beyond us.

2.f. Trajectory conflict detection

A new conflict detection method has been implemented and examined.

Previous versions of the software relied on checking the position of theaircraft at specific times into their future flight plan and detecting conflicts atthose points only. This method had many limitations.

In order to obtain more accurate conflict detection, where the positionwhere each conflict would start and stop would be recorded, we have



implemented a new approach based on the detection of solid intersections.Each flight’s position is checked against all other flights in the database for afuture time specified by the user. If two flights are close enough to present apotential conflict then they are checked in more detail. Pairs of line segmentsfor the pair of flights are constructed, with each line pair representing exactlywhere the flights will be at a certain time in their flight plans. These line-setsare then tested for those points where they become too close or where theyagain achieve sufficient separation and so we are able to trap the positionswhere the flights may enter and leave a conflict situation.

Any conflicts detected are presented to the user through a graphicalrepresentation of the start and stop points of each conflict and an audiowarning of the presence of a conflict using the positional audio approachdescribed above.

In this way the controller is led to the position of the conflict and can resolveit though interactive manipulation of the flight trajectory data.

This ‘brute-force’ detection approach is sufficiently fast on existing hardwareto provide interactive detection of conflicts as the controller manipulates theflight trajectories.

Figure 2-4 – Flights with existing or predicted conflicts are shown using colourbars along the planned flight path. These are updated in real-time as the user

modifies the flight plan so that the conflicts can be interactively resolved.



2.g. New Glove selection methods

In order to improve the selection methods we have attempted a newapproach based on a natural selection method using the data glove. Instead ofusing a wand pointer from the hand to the object of interest we nowconstruct a pointer by aligning the hand and the user’s dominant eye anddisplay a marker or cross-hair along this line.

This would be the natural way of aiming and pointing at objects in real life.When aligning things the user usually uses either the left or right eyeaccording to their personal eye dominance but in the stereo display systemthe dominance is usually overlooked, causing problems with distancemeasures for displaying the selection graticule.

Figure 2-5 – The user’s line of sight, through the glove position, is used tocontrol the position of the selection cursor which is visible in the centre ofthe screen in this figure. The user moves the head and hand to align thecursor with the object and carries out selection operations using finger

gestures. In essence the user just ‘grabs’ the object to select and move itproviding a very natural interaction metaphor.



Using this approach was often disturbing for a user of our system so weelected to follow an approach based on the dominant eye where thetargeting graticule is displayed in only one of the two ‘cameras’ of the stereodisplay.

This removes the problem of targeting and makes for a very natural selectionmechanism with the user’s hand both guiding the targeting graticule, invokingthe selection of the object using a hand ‘pinch’ and then moving the objectand releasing it entirely using their fingers.

One disadvantage that we have observed with this method is the fact that,with existing tracking and glove technologies, the weight of the equipmentcan become a problem since the user’s arm gets tired after a while if frequentselections and manipulations are required but the development of smaller andlighter tracked glove devices will remove this problem.

The incorporation of improved voice command selection would also reducethe number of occasions where the user’s hand selection must be employed.

2.h. Improved geographical orientation

In previous versions of the 3D VR ATM system we relied on the geographicalfeatures of the environment to provide orientation information to thecontrollers but some commented that this was insufficient, particularly whenzoomed in close to the scene.

We enhanced this by adding a ‘compass rose’, displayed in the top-rightcorner of the screen in an attempt to aid the controller in retaining a sense oforientation within the immersive scene.

This year we have employed a new approach where a compass rose is,instead, projected and blended into the map. When using the semi-immersiveworkbench display the compass is positioned with respect to the direction ofthe user’s head making it visible at all times.

The compass is scaled according to the distance between the map and theuser’s head providing the user with both orientation and ‘zoom’ informationas they work within the system.

We have also experimented with using a smaller circle, within the compassrose, which can be used to select flights and waypoints when combined witha suitable voice command set.



Figure 2-6 – The ‘ground compass’ image is blended into the display at thecentre of the user’s view direction in real time as the system updates so thatit is always visible in their current area of interest. This allows it to act also as

a selection mechanism when combined with a voice command.

2.i. Improved weather data representation

The previous versions of the system have included increasingly complexweather information which it is hoped will be of value to the controller incarrying out the task of flight management but the representations have been,necessarily, complex.

These representations have the problem that they could lead to aninformation overload situation where the user is swamped with informationand prevented from interpreting it quickly enough to carry out their task.

This previous weather display is shown in Figure 2-7 below.

Thus we have spent some effort this year in exploring ways in which it mightbe possible to compress the data presented to the user by identifyingparticularly significant features using techniques such as surface extractionwhich can be used to extract and present areas of extreme weatherconditions. The settings for this feature can be defined by the user while thesystem is running since these features are extracted in real-time from the rawdata loaded into the system.This new representation is shown in Figure 2-8 below.



Figure 2-7 – Weather display in the Phase III development system. Thiscomplex display includes a lot of different types of useful weather data buthas a tendency to confuse the user as much of the information it displays is

not significant.

Figure 2-8 – Areas of weather such as pressure systems or strong airmovement are automatically extracted from the raw weather data and

displayed using transparent surfaces. The transparency ensures that the flightinformation is still clearly visible to the controller.



3. Conclusions

This application continues to be developed according to methods that NVIS,as visualization and interaction experts, think appropriate to implement andtest. Each new feature raises new problems and new possibilities which wehope to pursue in future versions of the software.

While we have frequent interaction with professional controllers, boththrough our connection with Eurocontrol’s Innovative Research Group andthrough our own contacts with the Swedish Civil Aviation Authority, the newfeatures which we present here have not yet been part of any formalevaluation process with professionals in the sphere of air traffic control andmanagement.

This is an area which we hope to address in future development incollaboration with the INO group, and we hope to continue thisdevelopment process in the future to provide new features and explore theirusefulness for the ATC community.

Introduction and Evaluation of Selection-by-Volume Approach


INTRODUCTION AND EVALUATION OF

SELECTION-BY-VOLUME APPROACH

Nguyen-Thong Dang(1,2)

(1) EUROCONTROL Experimental Centre, Brétigny, France.

(2) Ecole Pratique des Hautes Etudes, Paris, France.

1. Introduction

This report presents the Selection-By-Volume approach, an approach forinteraction with three-dimensional (3D) objects through a combination ofgeometric shape and two-dimensional (2D) menu system. The main objectiveof this approach is to help overcome the difficulties in interaction withoccluded objects in a 3D environment. This approach also introduces adifferent way to create 3D interaction techniques. The report describes mainfeatures of this approach and also shows the results received from anempirical evaluation on two interaction techniques: Transparent Sphere andTransparent Cylinder, derived from this approach.

In an investigating study on the applicability and the usability of 3Dstereoscopic environment for Air Traffic Control [3], we found that theproblem of occlusion has considerable effects on the users’ performance ininteraction with 3D objects.

The problem of occlusion is originated by the relative location of theviewpoint of the observer in the 3D scene. In fact, given a certain viewpoint,some objects may appear either partially or totally occluded to the observer.Normally, before being able to interact with the occluded object, users haveto move themselves or to even change the topology of the scene in order togain visibility of the occluded objects. These disadvantages can be explained inFigure 1-1 and Figure 1-2.

Figure 1-1 shows an example of selecting objects using the wand as inputdevice and the Ray-casting [1] as selection technique. Users try to selecttarget object (the blue triangle) that is occluded by the yellow rectangle.Users have to move the wand to eliminate the occlusion state between theblue triangle and other objects.



A

A A

NO NO OK

Figure 1-1 – Change the point of view to avoid the occlusion state.

Figure 1-2 shows another way to select the occluded objects. Users changethe scale of the scene (e.g. by zooming) to eliminate the occlusion statebetween the blue triangle and other objects.

A A

zoom

NO OK

Figure 1-2 – Change topology by zooming to avoid the occlusion state

The handles to eliminate the occlusion state, as in Figure 1-1 and Figure 1-2,require several moves and controls, and cause tiredness to the users.

Although several interaction techniques have been proposed, only a smallnumber of these techniques have taken into consideration this problem.However, they do not offer a complete solution for the problem of occlusion.VooDoo Dolls [7] and Flexible Pointer [6] can deal with occluded objects butare quite complicated because they require two-handed interaction. The ray-casting technique [1] requires many manoeuvres to select the occludedobject as shown in Figure 1-1 and Figure 1-2 above. Virtual Hand and other



techniques based on Virtual Hand metaphor [8] were not designed to dealwith occluded objects. Thus, they cannot deal with the problem of occlusion.

Visual solutions can be used to help “disocclude” the occluded objects.However, the dynamic aspect of air traffic can still cause the occlusionbetween 3D objects. In addition, the visual solutions can cause the unnaturalperception of the traffic [9].

In an attempt to facilitate the interaction with occluded objects, we proposeda novel approach for object selection. The basic idea of this approach is touse a combination between the selection volume (under the form ofgeometric shapes) and 2D menu systems for interaction with 3D objects. Wecalled this approach the Selection–By–Volume approach.

2. Previous works

The idea of using a geometric shape for interaction with an object wasexploited in the past though cone-casting based technique such as Spotlighttechnique [5] and Aperture technique [3] and the Silk Cursor technique [10].

The Spotlight technique [5] uses a conic selection volume. All the objectswithin the conic selection volume may be selected. If there is only one objectin the selection volume, the user can trigger a command to select it. If two ormore objects fall into the selection volume, then the object that is the closestto the centre line of the selection cone is selected. If the angle formed withthe centre line of the selection cone is the same for both objects, the objectcloser to the device is selected.

The Aperture selection technique [3] is a modification of the spotlighttechnique that allows the user to interactively control the spread of theselection volume. By changing the selection volume, users can change theobjects that are selectable within the volume.

However, by using these two techniques, it is difficult to select a single objectin a set of closely spaced objects (as stated in [3]). Therefore, the Spotlighttechnique and the Aperture technique do not help overcome the difficultiesin the selection of occluded object.

Zhai’s silk cursor [10] selects objects that fall within a transparent cubicvolume. The transparent cubic volume provides visual cues that indicate thelocation of target objects that can be behind, within, and in front of thetransparent cubic volume. However, it is also difficult to select an object in aset of closely spaced objects or of occluded objects because users have todecide which part of which object is behind, within or in front of thetransparent cubic volume. This complicates the selection task.

The use of 2D menu system in combining to the volume selection in theSelection-By-Volume approach offers a disambiguation method for selecting a



single object in a set of closely spaced objects or of occluded objects. Thefollowing section introduces the main features of Selection-By-Volumeapproach.

3. Selection–By–Volume Approach

By using the Selection–By–Volume approach, the selection task can beperformed in two steps:

Pre-selection. Highlight objects (that include the object of interest)located within a predefined volume and

Se lec t ion . Pick up the target object by pointing at thecorresponding label (or menu item) in a menu that displays alllabels of objects, which fall within the selection volume.

The next sections discuss two main components of the Selection–By–Volumeapproach: the selection volume and the 2D menu system.

3.a. Selection Volume

The selection volume is characterized by the shape that represents theselection volume and by the size, the material of this shape.

Shape

Basic shapes such as cone, cylinder, sphere, cube etc. as well as other special,complex shapes could be used to represent the selection volume. Interactiontechniques created from this approach will be different from each other bythe choice of geometric shape for selection volume.

Size

The size of the selection volume decides the number of objects, which fallwithin the volume. This number decides the number of labels on the menu. Inthe same 3D scene, a large selection volume could contain more objects thana small one. In case of large selection volume, many objects can fall within thevolume, and it will take more time to select the object of interest. However,it’s easier to capture the target object with a large selection volume. On thecontrary, a small selection volume may require several handlings to capturethe target object. However, there will be fewer objects within the selectionvolume and it will take less time to select the object of interest. Therefore, areasonable choice of the selection volume size helps obtain an optimalnumber of objects, which can fall within the volume and thus improve theselection performance. Nevertheless, the number of objects located withinthe volume selection depends on the density of objects in the 3D scene aswell. With the same selection volume size, there could be more objects thatfall within the volume in a 3D scene with high density of objects than in a 3D



scene with low density of objects. Consequently, the size of the selectionvolume should be considered with regard to the density of the 3D scene.

Material

The material of selection volume is characterized by the color and thetransparency:

Color. The color is used to make the volume selection easy to see. Forexample, the use of color for the border of the selection volume could helpuser distinguish the parts inside and outside the volume selection.

Transparency. A specific degree of transparency for the selection volumeshould be considered in order to keep avoiding the situations in which theselection volume occludes or reduces the visibility of objects located withinthe selection volume.

3.b. 2D Menu System

Menu is one of the key components in the WIMP (Window, Icons, Mouse,Pointer) paradigm of the 2D user interface. In computer science, menu isknown as a list of options (menu items) from which the user can make aselection and afterwards performs a desired action. The use of menu in thiscase helps the selection of 3D objects become the selection of menu itemsarranged on a plan, which is easier and avoids the ambiguousness. Threeimportant features of menu design are the placement of menu, therepresentation of menu and the way to select items on the menu.

Placement

The placement of the menu influences the user’s ability to access the menuand the amount of occlusion of the environment (because menu can occludeobjects in the scene). In general, there are three types of menu placement:world-referenced menu (menu system that is placed at a fixed location in the3D scene), view-referenced menu (menu system is always placed at a fixedlocation compared to the viewpoint position) and object-referenced menu(menu system is attached to an object in the 3D scene) [4].

Representation

Orientation and style are among the key features of menu presentation:

Orientation. The extra depth dimension results in several possibilities forpositioning the menu in 3D scene. In fact, menus can be displayed on any XYplane, XZ plane, YZ plane or any other plane within 3D space. Anyhow, themenu should be positioned, oriented such that it is visible and easy tomanipulate.

Style. The commonly used menu styles in 2D user interface such as pop-upmenu, pull-down menu can be used for menu design in this approach.



Another feature of menu representation is the color of menu. In general, themenu color should make menu visible and should be designed with regard tothe color of the scene.

Item selection

Menu placement and menu presentation help visualize the menu beforeselecting an item on menu. In order to pick up the target object, the user hasto select its corresponding item on the menu. The item selection thus playsan important role for selection performance in this approach.

There is no more occlusion problem when selecting menu items. Thetraditional ray casting technique or the virtual hand can all be used for itemsselection.

3.c. Interaction techniques derived from theSelection–By–Volume approach

An interaction technique can be created from this approach by choosingfeatures for selection volume (i.e. shape, color, size etc.) and for 2D menusystem (i.e. style, menu placement, selection technique for items selection.).Detailed choices of each feature are mentioned above.

4. Evaluation of Selection–By–Volume approach

Based on the Selection–By–Volume approach, we created the TransparentSphere technique and Transparent Cylinder technique. These two techniqueswere used for the evaluation of the Selection–By–Volume approach, inparticular in dealing with occluded objects.

4.a. Transparent Sphere

The Transparent Sphere technique works in a similar way to the Virtual Handtechnique. Instead of using a hand-shape cursor to touch the target object forselection, a sphere volume is used. The sphere volume is transparent so thatusers can see the objects inside the sphere. Regarding the menu, an uprightpop-up menu attached to the sphere is used in items selection. This is a kindof object-referenced menus, which is activated after the pre-selection stepand disappears after having done the selection step. The ray casting techniqueis used to point to the items on the pop-up menu. An example of using theTransparent Sphere for selection is shown in Figure 4-1.



P1

P2

P3

P4P6

P5

P1

P2

P3

P4P6

P5

P2

P1

P3

P1

P2

P3

P4P6

P5

a) b) c)

Figure 4-1 – The Transparent Sphere technique a) The objects located withinthe transparent sphere are highlighted b) A ray is used to select a label (P3)

on the pop-up menu c) After being selected, the object P3 is in the centre ofthe sphere and is ready for manipulation task.

4.b. Transparent Cylinder

The Transparent Cylinder technique works in a similar way to the Ray-castingtechnique. Instead of using a light ray to touch the target object for selection,a cylinder volume is used. Similar to Transparent Sphere, the cylinder volumeis transparent so that users can see the objects inside the cylinder. Also, anupright pop-up menu attached to the cylinder is used for items selection. Thisobject-referenced menu is activated after the pre-selection step and isdisappeared after having done the selection step. The ray casting technique isalso used to point to the items on the pop-up menu. An example of usingthe Transparent Cylinder for selection is shown in Figure 4-2.

P1

P7

P3

P2

P8P5

P0

P9

P4

P6

P1

P7

P3

P2

P8P5

P0

P9

P4

P6

P1P7

P3

P2

P8P5

P0

P9

P4

P6

a) b) c)

P8

P1

P7

P2

P3

Figure 4-2 – The Transparent Cylinder technique a) The objects locatedwithin the transparent cylinder are highlighted b) A ray is used to select a

label (P1) on the pop-up menu c) After being selected, the object P1 is readyfor manipulation task.



4.c. Objective

The objective of the evaluation was to compare the users performance withthree different interaction techniques: the Transparent Cylinder technique,the Transparent Sphere technique and the ray-casting technique.

The subjects who took part to the experiment were engaged in someselection and positioning tasks.

Since the main hypothesis is that our proposal (the Selection-By-Volumeapproach) should support a quicker performance when the subjects have toselect partially occluded objects, the subjects engaged in the experimentperformed their tasks with both occluded scenes (in which the target objectwas occluded by other objects) and non-occluded scenes (where all theobjects were clearly visible). The Transparent Cylinder and TransparentSphere were involved in the evaluation at the same time to see the effect ofgeometric shape in this approach.

4.d. Experimental Design

The experiment was a within subjects design, with one independent variable(the interaction technique type) with three levels: the Transparent Sphere,the Transparent Cylinder and the Ray-casting. In order to avoid carry overeffects, the order among the three conditions was counterbalanced.

The reaction time of every task was recorded.

The materials of the experiment were composed as follows.

In all scenes, solid spheres were used as the objects to be selected and thenplaced in a specific point of the space. In all scenarios, the sphere P01 (ingreen color in Figure 4-3) was the object targeted for selection while thesphere P02 (in pink color in Figure 4-3) was target for positioning task. All thepositions of the spheres, including the green – or P01 – and the pink – orP02 – spheres, changed for every scene.

However, as it was mentioned before, half of the scenes displayed non-occluded target objects, while the other half displayed partially occludedtarget objects.

Therefore, for the occluded scenes, the place of the target sphere (the P01)was not randomly chosen, because its position had to be partially occluded.Thus, small adjustments on the spatial positions of the targets were done.

Every subject performed the tasks with 66 scenes (22 scenes * 3 interactiontechniques). However, only 48 objective measurements were taken, becauseevery subject was engaged in 18 training trails (6 training trails * 3 interactiontechniques). Half of the training trails comprised occluded targets and theother half non-occluded targets. The order of presentation of all the sceneswas randomized.



Figure 4-3 – Snapshots of experiments.

4.e. Procedure and tasks

A paper with written instructions was provided to the participants. They wereallowed to ask further explanations about the task, but only before thebeginning of the test. As mentioned before, every subject performed threesessions of trials: one with the Ray casting, one with the Transparent Cylinderand one with the Transparent Sphere.

Each session started with six non-measured trials. After that, the subject hadto perform sixteen measured trials: eight trials with occluded scenes and eightother trials with non-occluded scenes, whose order of presentation wascounterbalanced

Every subject had to perform a selection task followed by a positioning task.The tasks were as follows.

The selection task required the participants to select the target sphere(sphere P01) placed within the 3D scene. Then, they had to move theselected sphere from its initial position to the final position, indicated by thepink sphere (the sphere P02).

An automatic counter measured the time performances of every subject. Forthe selection task, the time was measured from the beginning of the test untilthe subject successfully selected the sphere (sphere P01). Then, anothercounter started: this counter measured the time used by every subject to



move the selected sphere to the target position (defined by the pink spherecalled P02).The task was finished when the sphere was placed to the target position.

Every scene was launched under the control of the subjects.

After the test, a questionnaire was provided to test participants. Questionsconsisted of 6-point Likert scale (ranging from strongly disagree to stronglyagree), but blank space was left after each question allowing free commentsand supplementary explanations.

When the subject had filled the questionnaire, an informal discussion aboutthe experiment was done.

4.f. Subjects

Twenty-four subjects (seven females and seventeen males) working atEUROCONTROL Experimental Centre were involved in the investigation.Their age average was 38.5, ranging from 22 to 61 years old.

4.g. Equipment

In order to prevent fatigue and strain, we arranged a small stage with a chairin front of the BARCO monitor, at a distance of about 50 cm. The subjectswere asked to sit while performing the task.

A test-bed was implemented using C++, Open Inventor version 3.11,CAVELib version 3.0.1 and a SGI Onyx2 workstation (512 Mbytes RAM). Thescenes were displayed using a BARCO projection system BARON (themonitor was 136cmx102cm). Subjects were using Stereographics Crystal Eyesglasses (120 Hz refresh rate) equipped with Intersense tracking system IS-900VWT. A six-degrees-of-freedom (6DOF) Intersense tracked wand withbutton and joystick was used as interaction device.

4.h. Experimental Results

Because the Transparent Cylinder and Transparent Sphere techniques aresolutions, which help users overcome the difficulties when dealing withoccluded scenes, we concentrate mainly on the experimental results withoccluded scenes. However, the experimental results with non-occludedscenes are also analyzed to show an overview about the performance ofTransparent Sphere and Transparent Cylinder in different cases.

The Friedman test will be used to see if there is any difference among threelevels (Transparent Cylinder, Transparent Sphere and Ray-casting). If there isno difference, we have the conclusion. Otherwise, a pair wise comparisonusing non-parametric Wilcoxon test will be applied to 3 pairs – (TransparentCylinder, Ray-casting), (Transparent Sphere, Ray-casting) and (TransparentCylinder, Transparent Sphere) – in order to see which technique have abetter performance in each pair.



First, the performed time (the sum of selection time and positioning time)was analyzed. Then, the analysis on selection and positioning time will befollowed. The Statistical is used for analyzing the experimental results.

Following sections analyze in order the occluded scenes, and the non-occluded scenes. The mean value is used for the Friedman test while themedian value is used for the Wilcoxon test.

Occluded scenes

Performed time. The means of the performed time are represented in Figure4-4. The mean value is 12.013 seconds (standard deviation: 2.779 seconds)for the ray-casting technique, 10.317 seconds (standard deviation: 2.355seconds) for the Transparent Cylinder technique and 9.820 seconds (standarddeviation: 2.103 seconds) for the Transparent Sphere technique. Thestandard deviations are shown in Figure 4-4 in parentheses.

The difference among three conditions is significant (Friedman test, p<.05).Therefore, a pair wise comparison is applied to 3 pairs: (Transparent Cylinder,Ray-casting), (Transparent Sphere, Ray-casting) and (Transparent Cylinder,Transparent Sphere). The median value is 11.987 seconds for the Ray-casting,9.877 seconds for the Transparent Cylinder and 9.575 seconds for theTransparent Sphere.

Occluded scenes

9,877

11,987

9,575

10,317(2,355)

12,031(2,779)

9,820(2,103)

0,000

2,000

4,000

6,000

8,000

10,000

12,000

14,000

median of performed time mean of performed time

Cylinder Ray Sphere

Figure 4-4 – Results of performed time with occluded scenes.

1. Transparent Cylinder, Ray-casting: the difference between the twoconditions is significant (Wilcoxon test, p<.05). Therefore, theTransparent Cylinder offers a better-performed time than Ray-casting.

2. Transparent Sphere, Ray-casting: the difference between the twoconditions is significant (Wilcoxon test, p<.05). Therefore, theTransparent Sphere offers a better-performed time than Ray-casting.



3. Transparent Cylinder, Transparent Sphere: the difference between the twoconditions is not significant.

Selection time. The means of the selection time are represented in Figure 4-5.The mean value is 7.721 seconds (standard deviation: 1.864 seconds) for theRay-casting, 6.007 seconds (standard deviation: 1.303 seconds) for theTransparent Cylinder and 6.285 seconds (standard deviation: 1.552 seconds)for the Transparent Sphere. The standard deviations are shown in Figure 4-5in parentheses.

The difference among three conditions is significant (Friedman test, p<.05).Therefore, a pair wise comparison will to be applied to 3 pairs – (TransparentCylinder, Ray-casting), (Transparent Sphere, Ray-casting) and (TransparentCylinder, Transparent Sphere). The median value is 7.874 seconds for the Ray-casting, 6.016 seconds for the Transparent Cylinder and 5.838 seconds forthe Transparent Sphere.

Occluded scenes

6,016

7,874

5,8386,007

(1,303)

7,721(1,864) 6,285

(1,552)

0,000

2,000

4,000

6,000

8,000

10,000

median of selection time mean of selection time

Cylinder Ray Sphere

Figure 4-5 – Results of selection time with occluded scenes.

1. Transparent Cylinder, Ray-casting: the difference between the twoconditions is significant (Wilcoxon test, p<.05). Therefore, theTransparent Cylinder offers a better selection time than the Ray-casting.

2. Transparent Sphere, Ray-casting: the difference between the twoconditions is significant (Wilcoxon test, p<.05). Therefore, theTransparent Sphere technique offers a better selection time than Ray-casting.


Positioning time. The means of the positioning time are represented in Figure4-6. The mean value is 4.310 seconds (standard deviation: 1.450 seconds) forthe Ray-casting, 4.310 seconds (standard deviation: 1.610 seconds) for theTransparent Cylinder and 3.535 seconds (standard deviation: 1.094 seconds)



for the Transparent Sphere. The standard deviations are shown in Figure 4-6in parentheses.

Occluded scenes

3,8594,132

3,345

4,310(1,610)

4,310(1,450) 3,535

(1,094)

0,000

1,000

2,000

3,000

4,000

5,000

median of positioning time mean of positioning time

Cylinder Ray Sphere

Figure 4-6 – Results of positioning time with occluded scenes.

The difference among three conditions is significant (Friedman test, p<.05).Therefore, a pair wise comparison will be applied to 3 pairs – (TransparentCylinder, Ray-casting), (Transparent Sphere, Ray-casting) and (TransparentCylinder, Transparent Sphere). The median value is 4.132 seconds for the Ray-casting, 3.859 seconds for the Transparent Cylinder and 3.345 seconds forthe Transparent Sphere.

1. Transparent Cylinder, Ray-casting: the difference between the twoconditions is not significant.

2. Transparent Sphere, Ray-casting: the difference between the twoconditions is significant (Wilcoxon test, p<.05). Therefore, theTransparent Sphere offers a better positioning time than the Ray-casting.

3. Transparent Cylinder, Transparent Sphere: the difference between the twoconditions is significant (Wilcoxon test, p<.05). Therefore, theTransparent Sphere offers a better positioning time than the TransparentCylinder.

Non-occluded scenes

Performed time. The means of the performed time are represented in Figure4-7. The mean value is 8.220 seconds (standard deviation: 1.896 seconds) forthe Ray-casting, 9.663 seconds (standard deviation: 1.926 seconds) for theTransparent Cylinder and 9.894 seconds (standard deviation: 2.744 seconds)for the Transparent Sphere.

The difference among three conditions is significant (Friedman test, p<.05).Therefore, a pair wise comparison is applied to 3 pairs: (Transparent Cylinder,Ray-casting), (Transparent Sphere, Ray-casting) and (Transparent Cylinder,



Transparent Sphere). The median value is 7.742 seconds for the Ray-casting,9.492 seconds for the Transparent Cylinder and 8.975 seconds for theTransparent Sphere.

Non-occluded scenes

9,492

7,7428,975

9,663(1,926) 8,220

(1,869)

9,894(2,744)

0,000

2,000

4,000

6,000

8,000

10,000

12,000

median of performed time mean of performed time

Cylinder Ray Sphere

Figure 4-7 – Results of performed time with non-occluded scenes.

1. Transparent Cylinder, Ray-casting: the difference between the twoconditions is significant (Wilcoxon test, p<.05). Therefore, the Ray-casting offers a better-performed time than the Transparent Cylinder.

2. Transparent Sphere, Ray-casting: the difference between the twoconditions is significant (Wilcoxon test, p<.05). Therefore, the Ray-casting offers a better-performed time than the Transparent Sphere.


Selection time. The means of the selection time are represented in Figure 4-8.The mean value is 4.312 seconds (standard deviation: 1.016 seconds) for theRay-casting, 5.645 seconds (standard deviation: 1.117 seconds) for theTransparent Cylinder and 5.860 seconds (standard deviation: 1.691 seconds)for the Transparent Sphere.

The difference among three conditions is significant (Friedman test, p<.05).Therefore, a pair wise comparison is applied to 3 pairs: (Transparent Cylinder,Ray-casting), (Transparent Sphere, Ray-casting) and (Transparent Cylinder,Transparent Sphere). The median value is 4.078 seconds for the Ray-casting,5.494 seconds for the Transparent Cylinder and 5.481 seconds for theTransparent Sphere.



Non-occluded scenes

5,494

4,078

5,481

5,645(1,117)

4,312(1,016)

5,860(1,691)

0,000

1,000

2,000

3,000

4,000

5,000

6,000

7,000

median of selection time mean of selection time

Cylinder Ray Sphere

Figure 4-8 – Results of selection time with non- occluded scenes.

1. Transparent Cylinder, Ray-casting: the difference between the twoconditions is significant (Wilcoxon test, p<.05). Therefore, the Ray-casting offers a better selection time than the Transparent Cylinder.

2. Transparent Sphere, Ray-casting: the difference between the twoconditions is significant (Wilcoxon test, p<.05). Therefore, the Ray-casting offers a better selection time than the Transparent Sphere.


Positioning time. The means of the positioning time are represented in Figure4-9. The mean value is 3.909 seconds (standard deviation: 1.071 seconds) forthe Ray-casting, 4.018 seconds (standard deviation: 1.232 seconds) for theTransparent Cylinder and 4.034 seconds (standard deviation: 1.405 seconds)for the Transparent Sphere. The standard deviations are shown in Figure 4-9in parentheses.

Non-occluded scenes

4,018(1,232) 3,909

(1,071)

4,034(1,405)

3,000

3,200

3,400

3,600

3,800

4,000

4,200

mean of positioning time

Cylinder Ray Sphere

Figure 4-9 – Results of positioning time with non-occluded scenes.



The difference among three conditions is not significant (Friedman test,p>.05). Therefore, there is no difference among three techniques in term ofpositioning time.

Questionnaire

The data derived from the questionnaires provide the following resultsrepresented in Figure 4-10. As previously stated, answers were provided usinga 6 points Likert scale.

Floating Menu. In the question regarded the “easy of use” of floating menu;the data account a mean value of 5.04. This shows a favorable opinion onusing floating menu.

Performance. In the question regarded the self-assessment of theperformance in term of selection speed, the data account a mean value of3.67 for the ray casting technique, 3.67 for the Transparent Cylinder and 3.96for the Transparent Sphere with no substantial difference between the threeconditions (Friedman test, p>.05). In the question regarded the self-assessment of the performance in term of test performing speed, the dataaccount a mean value of 3.38 for the ray casting technique, 3.63 for theTransparent Cylinder and 3.58 for the Transparent Sphere with no substantialdifference between the three levels (Friedman test, p>.05).

Usability. With questions regarded the self-assessment of the usability (easy touse), the data account a mean value of 4.04 for the Ray-casting, 4.33 for theTransparent Sphere and 4.08 for the Transparent Cylinder with no substantialdifference between the three levels (Friedman test, p>.05).

Enjoyment. With questions regarded the self-assessment of the enjoyment,the data account a mean value of 4.67 for the ray casting technique, 4.96 forthe Transparent Sphere and 4.67 for the Transparent Cylinder with nosubstantial difference between the three levels (Friedman test, p>.05).

Questionnaire

3,384,04

5,04

3,673,63

4,334,673,67

4,673,96

3,584,08

4,96

0,00

1,00

2,00

3,00

4,00

5,00

6,00

help perform quicker easy to use help select quciker enjoyable Menu is easy to use

Cylinder Ray Sphere

Figure 4-10 – Statistical results of questionnaire.



In all questions, the scores received from the Transparent Cylinder and theTransparent Sphere techniques are always better than the scores from theRay-casting. Although, statistically speaking, these differences are notsignificant; they also show a favorable opinion on using the TransparentCylinder and the Transparent Sphere.

Discussion

Occluded scenes. In case of occluded scenes, the Transparent Cylinder andthe Transparent Sphere can provide a better performance than the Ray-casting in term of performed time (i.e. the sum of selection time andmoving/positioning time). This better performance is based mainly on theselection time. In fact, both Transparent Sphere and Transparent Cylinderreport a better selection time than the Ray-casting.

In case of positioning time, the Transparent Sphere report a better time thanthe Ray-casting and Transparent Cylinder technique. There is no differencebetween Transparent Cylinder and the ray casting in term of positioning time.This shows that the sphere support users better than the cylinder in thepositioning task (in case of occluded scenes). This stresses also theimportance of choosing the geometric shape of selection volume for theinteraction technique derived from the Selection-By-Volume approach.

Non-occluded scenes. In case of non-occluded scenes, the Ray-castingtechnique is obviously better than the Transparent Cylinder techniquebecause the selection in Transparent Cylinder technique is performed in twosteps while the ray casting offers one-step selection. The experimental resultsalso show that the ray casting technique can provide a better performancethan the Transparent Cylinder and the Transparent Sphere technique in termof performed time and selection time.

Menu. The use of floating menu received a high score (5.04/6) and positivefeedbacks from test subjects. The floating menu should be considered as acomplement to existing 3D interaction techniques. In fact, the menu can beused to show different options for a manipulation task, or to help the existingtechnique solve the problem of occlusion.

5. General Discussion

The problem of occlusion.

A solution for the problem of occlusion as confirmed in the experimentalresults can be considered as the main advantage of interaction techniquecreated from the Selection–By–Volume approach.



Moving objects.

Thanks to the selection volume, the time that the target object stays insidethe volume could be long enough for performing the selection command.Therefore, with the selection volume, the users can catch the moving objectseasier than in the case of the ray-casting technique or the virtual hand.

Two-step selection.

The selection techniques created from this approach are indirect technique. Infact, the selection of object is performed within two steps and though menuselection. The direct selection techniques such as ray-casting and virtual handbased techniques can be more efficient in selecting the non-occluded object.

Number of objects in the selection volume.

If there are numerous items in the environment, the menu could potentiallybe long to ensure the user could quickly identify the objects. This menu couldthen occlude the objects. This issue should be taken into account in thechoice of the size of selection volume and in the design of 2D menu systems.

6. Conclusion

Through Transparent Sphere and Transparent Cylinder, theSelection–By–Volume approach was shown to be able to solve the problemof interaction with occluded objects. With the use of 2D menu system andthe selection volume, the Selection–By–Volume approach can be consideredas a novel approach for designing the interaction techniques in 3Denvironment. In addition, this approach also introduces the principles forcreating a new class of interaction techniques. The interaction techniquescreated from this approach are different by the choice of the selectionvolume, the 2D menu system and the way to interact with the 2D menu.

7. References

[1] Bolt, R., “Put-that-there”, In Proc. of SIGGRAPH 1980, ACM press, 1980,pp. 262–270.

[2] Dang, N.T., H.H. Le, and M. Tavanti, “A Multidisciplinary Framework forEmpirical Analysis of the Applicability of 3D Stereoscopic in Air TrafficControl”, In Proc. of IEEE CIRA’03, Kobe, Japan, July 2003.

[3] Forsberg, A., K. Herndon, and R. Zeleznik, “Aperture based selection forimmersive virtual environments”, In Proc. of the ACM Symposium onUser Interface Software and Technology, 1996, 95–96.

[4] Kim, N., G.J. Kim, C-M. Park, I. Lee, S.H. Lim. “Multimodal MenuPresentation and Selection in Immersive Virtual Environments” PohangUniversity of Science and Technology, Korea.



[5] Liang, J. and M. Green, “JDCAD: a highly interactive 3D modeling system”,In Proc. of Third International Conference on CAD and ComputerGraphics, Beijing, China, pp. 217–222

[6] Olwal, A. and S. Feiner, “The Flexible Pointer: An Interaction Technique forSelection in Augmented and Virtual Reality”. ACM Symposium on UserInterface Software and Technology (UIST '03), Vancouver, BC, 2003,81–82.

[7] Pierce, J., B. Stearns, and R. Pausch, “Voodoo dolls: seamless interaction atmultiple scales in virtual environments” SI3D 1999, pp. 141–145

[8] Poupyrev, I., M. Billinghurst, S. Weghorst and T. Ichikawa, “Go-GoInteraction Technique: Non-Linear Mapping for Direct Manipulation in VR”.In Proc. of the ACM Symposium on User Interface Software andTechnology, 1996, 79 – 80.

[9] Tavanti, M. “On the Relative Utility of 3D Interfaces”. Ph.D. thesis, UppsalaUniversity, 2004.

[10] Zhai, S., W. Buxton, and P. Milgram. “The “silk cursor”: Investigatingtransparency for 3d target acquisition”. Proceedings of CHI’94, pages459–464, April 1994.

Augmented Reality Tools for Tower Control


AUGMENTED REALITY TOOLS FOR TOWER CONTROL

Magnus Axholt

Complex System Modelling and Cognition European Joint Research Lab


1. Introduction

Augmented Reality (AR) systems combine three-dimensional computer-generated imagery with the view of real environment in order to makeunseen objects visible or to present additional information [4].

The application of AR within ATM is motivated because it displaysinformation that the controller cannot detect with her own senses; it amplifiesthe intelligence of the situation; it reduces head-down time and has the abilityto merge interfaces. Furthermore it will serve as an intermediary technologyand assert new interaction schemes for future applications where reality isreplaced for a virtual one.

One of the most basic problems and crucial enabling technologies for AR isthe matter of accurately tracking the user’s view and the objects in the realityseen by the user. This positional and orientational information is used by thegraphics rendering computer to correctly render the augmented information,foremost to position objects and adjust user’s perspective.

The computer-generated imagery that is superimposed on the user’s viewmust also correctly align with the reality or else the credibility and usability ofthe AR application will be reduced. The process of aligning reality with thecomputer-generated imagery according to position, and orientation of userand object, is known as registration [1].

Tracking equipment can currently be divided into five categories: Magnetic,acoustic, optical, mechanical and hybrid systems, all of which can be furthercategorized depending on what principle they use for tracking.

The tracking equipment can be evaluated and compared by a common set ofmetrics: Accuracy, resolution, update rate, latency and working volume.

If an AR system is used for an application where the metrics of the trackingsystem does not correspond to the needs, there will inevitably be registrationerrors. To be able to study this error, improve the tracking technology, oradjust the AR system, there is a need for a working error model andappropriate correction methods, all of which is dependant on the applicationof the AR system.



2. Related work

There have been several surveys of tracking techniques [3]. But, as technicalevolution progresses new tracking principles become available to us. Mergingthe strengths of various tracking systems into hybrid solutions allow tailoredsolutions to specific applications. Medicine is an area that has received a lot ofattention. Very little is done considering the needs of the ATM domain.

Registration errors in AR systems have previously been investigated byHolloway [4]. To estimate the error and make corrections for it MacIntyre etal. suggest a probabilistic approach [6] and also a measurement similar to thewell-known level-of-detail measurement in computer graphics, namely thelevel-of-error [5].

Refined tracking must not necessarily rely on more accurate measurementsaccording to the set metrics. Azuma has characterized two classes of head-motion predictions by analyzing them in the frequency domain determininghow accurate they are. He suggests an approach though which tracking errorcan be calculated given the characteristics of the user motion [2].

3. Formulating a PhD Thesis

The PhD studies started in November 2004 and the PhD thesis is stillundergoing iterations of literature studies, inducing assumptions, posingconstraints on problem space and subsequently a future solution space.

Figure 3-1 – Conceptual model of thesis formulation.

The current suggested topic is “Non-intrusive tracking for optimal registrationin ATM Tower Control AR Environment”.

Problem Space Solution Space

Constraints Assumptions

Literature Search

Augmented Reality Tools for Tower Control


Tracking, registration and AR are by their definition above. The remaininginterpretation of the suggested topic is discussed below.

3.a. Non-intrusive

Intrusiveness is hard to define since it does not necessarily need to be aquantitative objective measurement. Intrusiveness can originate from someequipment that introduces muscular or mental fatigue, that violates personalspace and integrity or that obstructs pattern of how user normally executes atask.

The challenge is to find a good definition and later to find a yardstick for thismeasurement.

3.b. Optimal

Optimal is normally not quantifiable by itself and needs to be set in a contextor evaluated with a reference point to make sense. The word optimal in thethesis proposal shall however be interpreted as the best choice of techniqueapplied in such a way that the most suitable tracking can exploit phenomenaderived from user’s practice. Parameters for achieving optimal tracking can bepattern of head movements, additional contextual information such as user’sposition in current task and the maximum tolerated level-of-error for a viewin the ATM Tower Control.

4. Future Work

Research and its accompanied experiments will initially be performed on aquite fundamental level. Since AR draws from many scientific disciplines,members of the 3D Research Team will jointly setup experiments andcontribute with their respective results.

The state of the art inventory will make it possible to categorize varioustracking principles. While evaluating these, it should be apparent if additionalparameters to the metrics are needed and, if so, how they should be defined.

Taking a task analysis into account, requirement specification for trackingequipment can be determined, given goal of task and its context. This will inits turn determine a yardstick for optimality. What is the problem and howcan it optimally be solved? It is important however to render results that areapplicable both within the ATM domain and in the AR field as a whole. Afundamental approach should provide this.

Examples of different approaches would be to investigate tracking techniqueseither by task and its context, by the possibilities given through hybridtechnologies or by applying entirely new concepts.



5. References

[1] Azuma, R., “A Survey of Augmented Reality”. SIGGRAPH 1995 CourseNotes #9 (Developing Advanced Virtual Reality Applications).

[2] Azuma, R., Bishop, G., “A Frequency-Domain Analysis of Head-MotionPrediction”. SIGGRAPH 95.

[3] Bhatnagar, D., “Position Trackers for Head Mounted Display Systems: ASurvey”, 1993.

[4] Holloway, R. “Registration Errors in Augmented Reality Systems”, 1995

[5] MacIntyre, B., Coelho, E., "Adapting to Dynamic Registration Errors UsingLevel of Error (LOE) Filtering” ISAR 2000. (Presented as a poster.).

[6] MacIntyre, B., Julier, S., Coelho, E. “Estimating and Adapting to RegistrationError in Augmented Reality Systems.” Proceedings of IEEE Virtual Reality,2002.

Visual Augmentation of Airport Objects in Air Traffic Control Tower


VISUAL AUGMENTATION OF AIRPORT OBJECTS

IN

AIR TRAFFIC CONTROL TOWERS

Stephen Peterson

Complex System Modeling and Cognition European Joint Research Lab


1. Introduction

Augmented Reality (AR) is the concept of enhancing (augmenting) the realworld. In comparison with Virtual Reality (VR), which completely immersesthe user in a synthetic environment, AR only supplements reality bycompositing real and virtual objects [1].

Effectively, in a visual context, AR is often achieved through computervisualization on a semi-transparent computer display device situated betweenthe user and an observed object in the real world. The observed real objectcan be overlaid with a computer generated virtual representation, taking intoaccount the positioning and orientation of both object and user. Thereby, thereal object and the virtual representation can be visible at same time user’svisual field. Efficient tracking technology is a fundamental criterion sinceaccurate spatial positioning and orientation data is required.

AR technology is especially suited for environments where users need torapidly retrieve and interpret additional data about an observed object, whileat the same time having visual contact with the real object and itssurroundings. If the additional information from different sources can bepresented in one single interface, and that interface is placed so that the usercan maintain visual attention of the surroundings, it could potentially help theuser to make better decisions in less time.

One such environment is the Air Traffic Control Tower, where its users, theTower Controllers, need to be able to observe the current traffic situation atthe airport visually. Simultaneously, a huge amount of additional information isavailable on different conventional computer screens located below thecontroller’s line of sight, sometimes referred to as Head-Down Displays,HDD [2].



2. Hypothesis

The hypothesis of this work is that airport object augmentation through ARdisplay systems in the Air Traffic Control Tower would help the controllerperform specific tasks. Potentially it could allow the controller to fasteridentify aircraft position and movement in all visibility conditions, while at thesame time not obstruct performance in other controller tasks. It could reducethe controller’s switching of focus between the inside (computer screens) andoutside (situation visible through the windows) of the tower, and therebyincrease the controller’s attention to important tasks.

3. Pre-Study

Initially, an exhaustive pre-study will be performed. Apart from an assessmentof the current Augmented Reality state-of-the-art, specific issues/problemsinherent to the Air Traffic Control domain will be studied. Optimally thecurrent and previous related work can be evaluated according to aclassification framework, for example Milgram’s Mixed Reality taxonomy [3].

3.a. Controller Tasks

In order to find proper applications of AR technologies in an Air TrafficControl Tower, a thorough study of the controller’s tasks must be performed.Specifically, the tasks concerning aircraft control on the airport and in itsimmediate vicinity are important. Some issues that need to be addressed are:

What decisions are made when controlling an aircraft? What tasks are considered particularly difficult or attention demanding?

Which sources of information are available to the controller?

How does the controller interact with existing systems in the tower?

How does the controller interact with pilots and other controllers?

When controller tasks have been analyzed and evaluated, assumptions can bemade on what tasks can be assisted by using AR technologies.

3.b. Enabling Equipment

A study of enabling equipment, specifically AR display equipment, will beperformed. Special attention will be taken to the following factors:

Intrusiveness

Stereoscopic (depth perception) capabilities

Collaboration possibilities

The display devices will be evaluated with regards to these factors along withthe suitability to the identified tasks in the controller task analysis.



4. Visual Augmentation Methods

Although the controller task analysis will address what precise tasks, if any,that could benefit from AR technologies, along with what display equipment isbest suited, some fundamental issues will certainly always have to beaddressed. In this section some initial ideas of visualization methods areproposed, specifically within the following three areas:

Aircraft Position and Orientation.

Aircraft Properties.

Affecting Phenomena.

This data is currently available to the controller only through conventionalcomputer screens, which requires the controller to divert attention from thesituation represented visually through the tower windows. How can thisinformation bepresented through an AR interface to the controller, so thatthe controller can consumethat information while visually observing theoutside situation?

4.a. Aircraft Position and Orientation

The concept of showing the position and orientation of an aircraft isintuitively a very important issue. Different visualization methods can beapplicable:

a. Overlaying the real aircraft with a photorealistic virtual representation

b. Overlaying the real aircraft with an outline of the aircraft’s shape

c. Showing a simple icon/shape indicating the position and orientation ofthe aircraft

d. Showing an icon indicating the position of the aircraft

e. Showing a large shape (e.g. sphere) that provides the approximateposition of the aircraft

These methods are more or less suited in different situations. For example,distant aircraft are hardly visible in reality, so outlining them as in a) or b) maynot enhance their appearance significantly. Furthermore, if the orientation ofthe aircraft is an important issue, simply indicating the position as in d) or e)will not suffice. Possibly different methods can be used simultaneously ifappropriate, depending on traffic density, distance to the objects, visibilityconditions etc. These questions will be addressed during the controller taskanalysis.

A fundamental aspect for the feasibility of the different visualization models isalso the availability of data. Some of the data required for the desiredvisualization methods may not be available at today’s airports. However, dataavailability will certainly change over time, as radars, satellite systems and



other navigational aids are improved. Therefore this work will assume thatappropriately accurate data is available with necessary refresh rates, as long asit is not completely unrealistic.

4.b. Aircraft Properties

Apart from aircraft position and orientation, other more abstract propertiescould be visualized. It would probably be advantageous to get informationabout the current velocity of an aircraft, and possibly its acceleration ordeceleration. Intuitively, velocities and acceleration/deceleration could bevisualized with pure numbers. Other options might be to indicate thequantities graphically, perhaps by showing an arrow with a length related tothe current speed or acceleration. Another option is to show a leader line,indicating the predicted path the aircraft will travel the next few seconds. Ifthis data is not available directly from a data source, velocity and accelerationapproximations could be derived from the positioning data mentionedpreviously by calculating the distance traveled in a known time interval (i.e.between two refresh events). Also, approximations of orientation could bederived in the same way. This is however only an extrapolation of historicaldata, and does not necessarily reflect actual conditions.

Furthermore it could be useful to see the altitude of an airborne aircraft(departing/landing) visualized. A possible approach would be to draw“droplines”, i.e. vertical lines connecting the aircraft to the ground. The lengthof the line gives a good indication of the altitude of the aircraft.

4.c. Affecting Phenomena

Data representing more abstract phenomena that greatly affects aircraftmanagement/airport operations would be interesting to visualize in the sameAR interface.

Weather data could be visualized in a number of different ways. For example,wind strength and direction could be shown with a vector field, where thevector length corresponds to the strength of the wind.

Similarly, the phenomenon of wake vortex turbulence could be visualized,provided that necessary data is extracted from a wake vortex sensor (radar)system at the airport. Also, since it is known that aircraft have a set separationdistance because of this phenomenon, this distance could be visualized as alag line (opposite to the leader line described above).

5. References

[1] Azuma, R T. A Survey of Augmented Reality. Presence: Teleoperators andVirtual Environments 6, 4 (August 1997), 355-385.



[2] Fränzl, T. Augmented Vision Information Systems for Air Traffic ControlTowers. International Conference and Workshop: Telecommunicationsand Mobile Computing, Graz (Austria), October 2001.

[3] Milgram, P., Kishino, F. (1994): A Taxonomy of Mixed Reality VisualDisplays. IEICE Transactions on Information and Systems, E77-D(12),pp.1321-1329

Air Traffic Tower Controller’s Tasks Performance in Augmented Reality Environment


AIR TRAFFIC TOWER CONTROLLER’S TASKS PERFORMANCE

IN AUGMENTED REALITY ENVIRONMENT

Elzbieta Pinska


Ecole Pratique des Hautes Etudes, Paris, France.

1. Introduction

The proposed studies are the continuation of multidisciplinary framework forthe empirical analysis of the applicability of 3D stereoscopic visualization andinteraction for ATC environment carried out at EUROCONTROLExperimental Centre. The framework composes of three components:

Human Factors,

Visualization and

Interaction.

The previous studies focused on the empirical investigation of memory taskperformance for 2D and 3D displays for ATC. The results show thatperformance using 3 D displays is task dependent [13][11].

Current studies concern evaluation of augmented reality visualization andinteraction for tower control traffic. The main attention will be given toperformance parameters assessment for tower control activity.

Keywords: human performance, augmented Reality, 3D, depth perception.

2. Background

The tower control activity is based on spatial orientation and integration ofthe airspace among the three dimensions. Controllers relay on the windowview visibility that depends on day/night time and existing weather conditionsor the radar that presents only two dimensions and where only a mouseinteraction is allowed. The controllers maintain three-dimensional mentalpicture of the traffic situation supported by these resources. Its is expected that augmented reality tools for control tower takes anadvantage of containing depth cues and can represent data in an easilyaccessible way for controller. The anticipated advantage of augmented realityis a realistic representation of picture regardless to outdoor conditions. Therepresentation includes realistic visual images of the objects (aircraft type),



additionally provides flight related information attached to the object likeaircraft tracks, vertical speed, flight plans and all tools that support controllerin his task. Presented scenarios offer high level visibility, apart from day/nighttime and also the weather events, like fog or clouds, might be presentedwithout screen saturation and occlusion.

Furthermore applying augmented reality displays acquires spatial manipulationand navigation among the representation. Also interactive Human MachineInterface (HMI) allows controller to act more flexible and suitable to hispreferences. Rapid and dynamic interaction, zooming certain spot of traffic orrotation that changing point of view might bring benefits for controllers,however further research are require to examine the most profitable use ofthis functions.

3. Tower Controller task description

In most of the cases, the Air Traffic Tower controller is responsible forseparation of aircraft traffic operating within 40 miles radius from the airport,dependently on letter of agreement [8].

There are different tower control positions, dependently on the area andresponsibility, called: Flight data Controller, Ground controller, LocalController, Approach and departing Controller (TRACON) and Clearancedelivery controller. Presented study concerns the local controller and s/he’stask performance. The local controller duty is to safely sequence arrivals anddepartures at the airport and to ensure the proper runway separationbetween aircraft. The local controller issues the appropriate instructions forto arriving and departing aircraft.Summarizing the main local controller duties are:

Determining the active runways

Issuing landing and takeoff clearances

Issuing landing information

Sequencing landing AC

Coordinating with other controllers

Issuing weather and NOTAM information to pilots Operating the runway and approach systems

4. ATC Tasks Performance

A task is defined as a sequence of single movements or mental operationsthat are allocated to the same goal and can be derived from the samegoal [5]. The experiments tested human performance in the environmentcontained depth cues provided inconsistent results.



The previous studies showed that effectiveness of 2D versus 3D stronglydepends on the task that was performed [9][13]. The performanceassessment in words of timing and accuracy for identifying flights levels wassuperior for 3D displays [11]. According to [6], adding the vertical profile toexisting horizontal plan view display has been proved to be quite successfulfor the task of vertical separation.

[9] studies show that compressing a 3D display into the perspective but stillflat display might create ambiguity along the line of sight for the viewer. Theambiguity might be more difficult to overcome for the viewer than theambiguity of depth information in 2D displays because in 3D it is spreadamong 3 axes. Also, it appeared harder to judge exact distances and angleswith 3D.

Ch. Wickens, in his experiments [13], shows lower cognition interaction ofinformation that means less effortful creating three-dimensional mental modelof the situation based on displays provided depth cues.

Another problem revealed in studies was visual attention for augmentedreality.

Experiment of Furstenau [3] showed the problem of switching attentionbetween the far and close visual domain. The other problem deals with user’sfocal attention point that is drawn to the centre of the screen: during thezooming of certain spots of traffic, other relevant parts of view should not beexcluded from controllers attention [4].

Presented results show the variety of results gained for three-dimensionaldisplays. All results appear to confirm the assumption that performance isstrongly task depended.

5. Augmented reality environment

The Augmented Reality system supplements the real word with virtualcomputer generated objects that appear to coexist in the same space as thereal word.

Augmented reality might be defined as [1]:

Combining real and virtual objects in real environments. Running interactively and in real time.

Register real and virtual objects with each other.

Augmented reality might be applied for tower control for as a support toolallowing augmentation and positioning of physical objects: aircraft, groundvehicles, buildings, runways and taxiways under all weather conditions.Computer generated visualization of weather condition and aircraftphenomena, the status of runway, taxi paths, without occluding the view, let



controller to merge information from different recourses in single interface.Additionally, it is reducing head down time and gives the possibility ofsupporting collaborative work.

6. Experiment design

The proposed study involves comparison of the particular tasks performancebetween the existing status AT Tower control and new environmentsupported by augmented reality tool.

The assessment of performance for the experiment will be based onfollowing characteristic:

Time of task execution

Accuracy of execution

Special attention will be given to the depth perception, distance judgment andspatial orientation issues.

7. References

[1] Azuma, R Baillot, Y. Behringer,R. Feiner, S. Julier,S. MacIntyre,B. “RecentAdvances in Augmented Reality”. IEEE Computer Graphics andApplications, 2001.

[2] Dang Nguyen, H. Le-Hong, M. Tavanti. “Empirical Analysis of theApplicability of 3D Stereoscopic in Air Traffic Control”. In Proceedings ofIEEE 6th ITSC2003, International Conference on IntelligentTransportation Systems, Shanghai (China), October 2003.

[3] Furstenau, N. Michael Rudolph, Markus Schmidt, Bernard Lorenz, DLRThorsten Albrecht, technical University of Braunschweig, Institute ofPsychology “On the use of transparent rear projection screens to reducehead down time in the air traffic control tower” Proceedings of Humanperformance, Situation Awareness and Automation Technol. Conf.(HPSAAII) 22-25.3.04, Daytona Beach, Florida (USA).

[4] Grunvald A, Shaviv G. “Advanced interactive display formats for terminalarea control”. Faculty of Aerospace Engineering, Technion, Israel,Institute of Technology, Haifa (Israel), 1999.

[5] Integrated Task and Job Analysis of Air Traffic Controllers - Phase 3: BaselineReference of Air Traffic Controller Tasks and Cognitive Processes in theECAC Area, EUROCONTROL HUM.ET1.ST01.1000-REP-05.

[6] Merwin, D. O’BrienJ, Wickens. “Perspective and coplanar representation ofair traffic, Implication for conflict and weather avoidance”. In proceedingsof the 9th Symposium of Aviation Psychology, pp 362–367, Columbus,Ohio (USA), 1997.



[7] Mooij,M. Dejkker,S. Weikert. C. “The future of Air Traffic control inSweden”, Vinnova, rapport VR 2001;15.

[8] Nolan, M.S Fundamentals of Air Traffic Control,1999, Brooks/ ColePublishing Company, USA.

[9] St John, Cowen, Smallman, The Use of 2D and 3D Displays for shapeunderstanding versus relative position task, Human Factors, Spring 2001,pp. 79–98.

[10] Smallman, H S, St John , Oonk H M , Cowen M. “Track RecognitionUsing Two Dimensional Symbols or Three Dimensional Realistic Icons”.Technical report, 2000 SCC San Diego, CA 92152-5001.

[11] Tavanti, H. Le-Hong, T. Dang-Nguyen, "Three-dimensional Stereoscopicvisualization for Air Traffic Control Interfaces: a preliminary study". InProceedings of the IEEE "22nd Digital Avionics Systems Conference",Indianapolis, Indiana, October 2003.

[12] Tham, M., & Wickens, C.D., 1993, Evaluation of perspective andstereoscopic displays as alternatives to plan view displays in air trafficcontrol, Technical Report (ARL-93-4/FAA-93-1). University of IllinoisInstitute of Aviation, USA.

[13] Wickens, C. D., 2000. The when and how of using 2-D and 3-D displaysfor operational tasks, In Proc. of the 44th Annual Meeting of the HumanFactors and Ergonomics Society, pp. 403-406. Santa Monica, CA:Human Factors Society.

Vital – Advanced Time-Line Approach for Future ATM Environments


VITAL – ADVANCED TIME-LINE APPROACH

FOR FUTURE ATM ENVIRONMENTS

Horst Hering


1. Introduction

The number of aircraft will increase in the future. It is commonly agreed thatin several high density traffic areas like central Europe the capacity limits arenearly reached. Predictions see in 4-Dimensional (4D – x, y, z-coordinatesand time) Air Traffic Management (ATM) a solution. As no revolution in ATCwill take place, the close future 4D ATM system will be human centred. Thehuman controller will still have to construct a mental picture of the air trafficfor his own understanding. This mental picture is required for anticipating andpredicting the future movements of the aircraft. Considerable mental effort isrequired. Human’s mental resources limit the number of aircraft handledsimultaneously in a sector.

The EUROCONTROL strategy paper [1] and a MITRE Corporation study [2]propose to introduce 4D ATM systems to increase capacity. Avionicindustries are in line with this vision, as in the 4D flight management systemsfor the cockpit reality. It is obvious that such a future 4D ATM systemrequires more complex information and it will includes new control conceptswith new features.

2. The Vital Artefact

Presenting this complex 4D information with current methods will increasethe mental workload of the controller. To overcome such constraints forfuture 4D ATM, a novel method is proposed to presenting this informationto the en-route controller in form of dedicated table. In this table each aircraftis represented by an artefact that may be placed to controller’s convenienceinside the table. The artefact, see Figure 1, content in digital and analoguerepresentation flight plan and real-time radar data. Flight plan and radar dataare permanently correlated and pre-processed for the analoguerepresentation on the base of a time-line. The time-line is representing alinear watch progressing in real-time. This animated time-line concept inspiredthe name Vital for this innovative concept.



Figure 1 – Vital Aircraft Artefact.

3. Vital Time Line Technique

Vital’s innovative technique is based on the principle of presenting flight plandata on a time-line. This technique was proposed in the early 1970s by Nobeland Sperandio [3] for en-route centres. Vital extend this idea with real-timeradar information correlated with the flight-plan data.

The project SuperSector of the EEC used a similar concept called DynaStrips.Early human factor evaluation results have been presented by Grau et al [4].In conclusion, they stated: ‘DynaStrips presents data to the controllers in aform which enables them to construct a more relevant mental image of theair traffic in a shorter time. By facilitating the controllers’ mental image, itallows them to work with greater anticipation, making it possible to manageheavy workloads more easily and safer.’

Hypothetically Vital’s time-line approach will have similar benefit from thedigital/analogue representation of data to the en-route controller. Vital couldbe an easy, natural, self-explaining interface for current and advanced ATMconcepts.

Vital gives a logical mathematical answer by applying known time basedalgorithm to create a analogue representation of pending uncertainty of theactual ATM system to support controllers perceptive and cognitiveunderstanding of the situation. That approach facilitates humansunderstanding and represents a strong cognitive help in form of a preliminarytreatment of available information for the operator.

The time-line is moving permanently in real-time to the left. A fix indicatorline is indicating the actual time. Displayed information reaching the left edgeof the time window disappears while on the right new information is filled up.Beacons and waypoints of the flight plan are shown in their chronologicalorder at the estimated time. These logical mathematical time estimations arebased on permanently updated radar data. Figure 2 shows an example of thetime-line data field of an aircraft artefact.

The identification of conflicts with Vital time-line field will be done by thecomparison of vertical alignments of current and extrapolated future aircraftpositions over the navigation points represented on the time-line, seeFigure 3. Aircraft flying the same level (or crossing) may conflict by:



Opposite traffic: two time-lines contain the same beacon/waypoint names,one is in reverse order, and there is a vertical alignment/overlapping of acommon segment, see Figure 3, time-line 1, 2.

Merging: the same single beacon/waypoint name is vertically aligned ofcrossing flight plan artefacts on the Vital tool, see Figure 3, time-line 3, 4.

Over speeding; flight plan artefacts with the same beacon/waypoint namesof a route segment, but with different time intervals on the time-line andthe vertically alignment of common segment, see Figure 3, time-line 2, 3.

DOM HAM OSN TOSPA

10:10 10:20 10:3

Actual time reference (fix)

Time-line (minutes) moving towards left

Beacons, Waypoints moving with time-line

Figure 2 – Vital’s time-line field.

Figure 3 – Conflict identification by vertical alignment of common time-lineinformation introducing colour for sector boundaries.

4. Vital in a 4D Environment

Future ATM is going toward a 4D navigation of the aircraft. Vital with its time-line approach could help the controller to reduce the complexity of the 4Dtrajectories for human mental understanding and supporting his prediction.



Due to this time-line, Vital is especially well adapted to support humans 4Dmedium/long term conflict detection and resolution in a graphical way. In afuture 4D ATM environment Vital could support all innovative R&D conceptslike route offsets, Airborne Separation Assurance System (ASAS), speed control,delayed or locally fixed climb/descend including an envelope of uncertainty.Vital could act as natural interface for the up linking of ATC instructions viaController-Pilot Datalink Communication (CPDLC).

Route offset

Aircraft on offset fly on a line parallel to the route with a fixed distance of e.g.5 NM .The offset may be right or left hand of the route. Similar to aircraftposition lights, red dots of the time-line indicates a left offset and green a rightoffset. Dragging a time-line dot or square slightly up or down, pops up awindow to select the offset from this time on.

Figure 4 – Time-line presenting route offset.

Climb/descend

Known flight level change of the aircraft are indicated by the Vital tool. Adiagonal blue strip indicates the estimated level change area. The differenttime length of the strips are related to uncertainty of the manoeuvredexecution. The angle of the strip is depending from the rate of climb/descendand the number of FL to move. The final Cleared Flight Level (CFL) is indicatedat the right end of the strips.



Figure 5 – Time-line presenting climb/descent.

Station keeping

Vital is able to indicate a station keeping ‘train’ of aircraft in a natural way onthe time-line, see Figure 6. The red bar represents the time segment on thetime-line which is attributed to the ‘train’ of aircraft. In the example the 3aircraft occupy about 8 minutes of time space. The controller handles this‘train’ as one unit which is represented for him by the first aircraft in line.

Figure 6 – Time-line presenting station keeping aircraft.

Speed envelope

Vital gives the possibility to indicate possible time envelopes over the nextwaypoints of the time-line. These earlier or later arrival of the aircraft overfuture waypoints are based on estimated safe speed variations (related toaircraft type, flight level). Figure 7 shows an example with ‘speed’ clicked.

Figure 7 – Time-line presenting speed envelope.



5. Conclusion

Vital proposes to improve the controllers’ mental image of the en-routetraffic by the innovative combination of representing digital and analogueaircraft data in a same artefact. The analogue data are extracted from radarsensors, correlated with flight plan information and represented pre-processed on a time-line.

An early evaluation of a time-line concept for strips showed benefit for thetime needed to create the mental traffic picture. Hypothetically similar benefitcould be expected by the concept of Vital.

Several new ATM concepts for a future complex 4D ATM can be supportedby Vital with a simple, intuitive and natural representation of the information.The Vital time-line approach may be adapted to current concepts too.

The EEC will realise the proposed ideas of this paper as a demonstrator userinterface of Vital with a graphical rapid prototyping tool. This paper describesits functionality for an early look and feel evaluation. A hypothetical En-routeMONitor (EMON) presenting Vital’s concept is shown in Apendix I.

6. References

[1] EUROCONTROL, 2004, Air Traffic Management Strategy for the Years2000+, EUROCONTROL, Brussels, Belgium.

http://www.eurocontrol.be/dgs/publications/brochures/v2_year2000_en/p7.html

[2] Mohleji, Ostwald, Sept. 2003, Future Vision of Globally HarmonizedNational Airspace System with Concepts of Operations Beyond Year 2020,The MITRE corporation, McLean, USA.

http://www.mitre.org/work/tech_papers/tech_papers_03/mohleji_airspace/index.html

[3] Nobel & Sperandio, 1973, Etude expérimentale du strip ‘base temp’ al’usage des centres des contrôle régionaux, Centre d’Etude de laNavigation Aérienne, France.

[4] Grau, Nobel, Guichard, Gawinowski, 2003, ‘DynaStrips’: A time-lineapproach for improving the air traffic picture of ATCOs, Proceedings of22nd DASC Conference Indianapolis, USA.



7. Appendix

How a EMON - Vital could looks like; an example constructed for theMunster-sector (MN) of the EUROCONTROL UAC Maastricht with mail-and recycling-box.

Figure 8 – EMON-Vital.

Wheelie – Mobile Horizontal Display Filter to Ease Controller’s Separation Task.


WHEELIE – MOBILE HORIZONTAL DISPLAY FILTER TO EASE

CONTROLLER’S SEPARATION TASK.

Horst Hering


1. Abstract

Filter techniques for horizontal FL layers are widely used for the ODS. Theaim of these conventional techniques is to reduce the displayed informationto the required level. This level of information is still complex (3D) andrequires strong mental effort for the separation task.

Wheelie introduces the concept of mobile filtering for a reference FL,accessed with the wheel of the mouse. Aircraft associated with the referenceFL are highlighted with an aureole. These aircraft may not conflict with anyothers that are not on the reference FL. Aircraft on the reference FL have tobe separated horizontally. So the sub-separation task is reduced in complexityfrom a 3D to a 2D problem. Therewith Wheelie stimulates the controllerwith a restricted ‘vision’ of his traffic and clue for a simpler ‘horizontallyseparated’ – yes/no answer.

2. Introduction

In the controlled airspace, safe aircraft separations have to be guaranteed bythe responsible controller of the sector. For safe separation the controller hasto apply horizontal or vertical separation. Conventional radar displays calledOperational Display Systems (ODS) represent the information in 2Dimensions (2D). With such a display, the horizontal separation betweenvarious aircraft is easily perceptible by the human operator. An experiencedcontroller is able to observe horizontal separation at first glance.

In case the horizontal separation is no more guaranteed, vertical separationhas to be applied. Vertical separation is based on flight level (FL) datacollected from the secondary surveillance radar. This FL information is shownas a number in the label associated with the aircraft symbol. The CoReHMI [1] specifies for these labels a minimum of two lines going up to six lines.Such a large number of label lines permits advanced ATC systems movingtowards a stripless environment. Currently, most conventional ATC systemshave the minimum two lines label as standard.



To observe safe vertical separation the controller has to scan all tracks underhis responsibility permanently. He has to read the actual FL values from thelabels, memorize them and compare them with each others to create amental traffic representation. This task requires strong mental effort from thecontroller and becomes more difficult in a stripless environment. Forsimplification, this paper shows in its examples minimum 2 lines labels, only.

3. The Idea of Wheelie

In 1999, Hering [2] proposed the idea of a horizontal filter for an ATC utilitycalled Mosaic. The presented idea exploits the approach of Mosaic: ‘aircraftflying on different FL can not cause horizontal separation problems’. As aconsequence, at first glance the controller can identify aircraft that are flyingon the same FL. All other cannot conflict with the selected aircraft. AbsoluteFL values are not really required to ensure vertical separation.

A horizontal filter function displays all aircraft of a selected FL differently onthe ODS than all others. The selected FL represents the reference for thehorizontal filter function. The reference FL is selected and changed easily withthe wheel of the computer mouse, therefore the tool is called Wheelie.

4. Wheelie’s Technology

The radar sensors provide the controller with much more information thanhe needs for his specific task. These sensor data are reduced by filters,selected by the controller. Therefore an ODS displays filtered aircraft, only.Further vertical filtering was proposed by H. David [3][4] combined withcolor coding to reduce complexity of the en-rout traffic display. Figure 1shows a small snapshot of a conventional en-route traffic situation. Actualfilters work as pre-settings for the ODS. Wheelie is designed for permanent,dynamic use. Scrolling the mouse wheel selects the reference FL. Aircraftflying the reference FL are highlighted in a graphically emphasis manner. -Wheelie never suppresses information shown on the ODS screen. – At firstglance the controller can identify potentially conflicting (horizontally) aircraftflying the reference FL. All other aircraft are flying on different FLs, they areout of conflict with the aircraft selected by Wheelie, regardless of theirhorizontal position.

Wheelie displays the selected reference FL number in a unobtrusive mannerin the background of the ODS. Aircraft flying on the reference FL get a virtual,invisible source of light behind the aircraft symbol ‘switched on’, which createsan aureole around the symbol. This idea was developed and evaluated by M.Tavanti [5].

The wheel of a mouse represents the ideal tool for changing the reference FL.The movements up and down with the wheel are natural to humans. In



general they need no supplementary training or mental effort to scroll themouse wheel. Scrolling all reference FL will be very quick, as a dozen of FLare used in an en-route sector, mainly.

Figure 2 shows the same traffic situation as in Figure 1, with Wheelie set tothe reference FL 330 and obviously the pending lost of separation, betweenAFR1304 and BAW2330, in about 2 minutes is stressing the observer.

For limiting the influence of Wheelie on the ODS image, Wheelie works ondemand only. Mainly Wheelie is sleeping and wakes up by a turn of themouse wheel. After the controller stops scrolling (delay i.e. 10…20s), theWheelie functionality falls into sleep again and disappears from the display.

TAP1104 330

BAW2330 330

KLM304 340

SAS433 360

AZA1004 360

IBE1122 310

DLH1394 320

AFR1304 330

AUA145 320TAP1104

330TAP1104 330

BAW2330 330BAW2330 330

KLM304 340KLM304 340

SAS433 360SAS433 360

AZA1004 360AZA1004 360

IBE1122 310IBE1122 310

DLH1394 320DLH1394 320

AFR1304 330AFR1304 330

AUA145 320AUA145 320

Figue 1 – OSD En-Route Snapshot.



Figure 2 – ODS with Wheelie: selected reference FL 330.

5. Basic HF aspects of Wheelie

Wheelie is neither another support tool to create controller’s mental trafficrepresentation, nor another safety-net tool. Wheelie will stimulate thecontroller to see his traffic situation under another, a restricted ‘vision’ relatedto the reference FL. These highlighted aircraft are displayed in a commonway, with an aureole to be distinguished from other aircraft flying on differentFL. Human factor ‘Gestalt’ principles let perceive the selection as a unit with acommon property, i.e. same FL. To this selection, the simpler horizontalseparation rules have to be applied. Horizontal separation demands from anexperienced controller less mental effort than vertical separation.

Wheelie’s user interface is based on the conceptual model of the user task,i.e. separation. The conceptual model is based on the internal representation,understanding and decision-making of the human, cf. IBM [6]. Wheeliestimulates humans perception with a part of the complete situation and cuefor an answer (separated ?) of the presented stimulus.



For Mandel [7] using real-world metaphors is one of the basis for userinterface design. For simpler human understanding, ATC organizes theairspace in a FL system, similar to floors in a shelf. The ODS superposes thesevirtual ‘floors’ to the known, complex single pane radar picture. Wheelie usesthe ‘shelf floor’ metaphor, see Figure 3. Wheelie filters the traffic in a way thata human can access a single virtual ‘shelf floor’, i.e. reference FL, to simplify thecomplex separation task. Scrolling the mouse wheel lets the controllernavigate onto the virtual ‘shelf floors’ as by a lift.

Figure 3 – Wheelie: Real World Metaphor.

Wheelie focuses the human attention to produce an answer to this restricted2D situation. The answers on these 2D cues will demand much less mentaleffort than the 3D situation. Manipulating Wheelie with the finger can be seenas routine task in the cognitive sense. Routine tasks, like body movements arecontrolled from humans lower level memory, not affecting humans workingmemory. So, supplementary mental effort by using Wheelie may not beexpected.

6. Conclusion

Wheelie’s user interface is based on the conceptual model to supporthumans internal representation, understanding and decision-making. It usesthe real-world metaphor ‘FL shelf’, which is the base of humans mental trafficorganization.

Hypothetically Wheelie may affect workload and situation awarenesspositively, but it will not affect the actual safety and security level in a negativesense at least. For an early ‘look and feel’ evaluation of Wheelie, the EECdeveloped a graphical user interface with a rapid prototyping tool. Apreliminary human factor study shall evaluate the potential benefits ofWheelie.



7. References

[1] EEC-ECHOES, 2004, A Human-Machine Interface for EnRoute Air TrafficControl - CoRe HMI Specification, EUROCONTROL ExperimentalCentre, France.

[2] Hering, 1999, Application de visualisation avancée 3-D pour unenvironnement de travail de contrôleurs aériens future, respectant lesfacteur humains, Université René Descartes, U.F.R Biomédicale, Paris,France.

[3] David, 1997, Radical Revision of En-Route Air Traffic Control, EEC-ReportNo. 307, EUROCONTROL Experimental Center, France.

[4] David & Bastian, 2001, Initial Evaluation of a Radically Revised En-Route AirTraffic Control System, EEC-Report No. 360, EUROCONTROLExperimental Center, France.

[5] Tavanti & Flynn, 2003, TRAMS, Visualizing Mode S Aircraft in ITI,Proceedings for HCI International2003, Crete, Greece.

[6] IBM, 1992, Object oriented Interface Design: IBM Common User AccessGuidelines, New York, USA.

[7] Mandel, 1997, The Elements of User Interface Design, Wiley, New York,USA.


ATMModelling

ATM Modelling


Major cause for delay in airtransport is assumed to come fromoperational uncertainty whoseimpacts are propagated throughspace - the network of adjacentsectors, and control centres, andtime – the network of adjacentcontrollers’ tasks. This research axisfocuses on analytical modelling ofATM natures and operations forbetter understanding the stochasticnature of ATM operations, fromATFM planning to ATC tacticalactions. The ultimate goal consists inbetter filling the gap between thepredictive components of ATFMand the adaptive components ofATC actions. Two major researchfocuses are:

Understanding and MasteringUncertainties through researchon models to capture thestochastic nature of ATM, and

Human Cognitive subsystem inATM system to capture theimpacts of schematic humancognition in safety managementobjective of ATM, as well as tobe able to perform morerealistic model-based ATMexperimentations.

All studies target the uncertaintynature of ATM from either an

empirical analysis or a theoreticalexplanation of the nature of theproblems; and fall in either theACARE SRA challenges or needs tooptimise the use of existing ATM-Airspace capacity or to maximiseairport performance.

On the research tackling the issuesof uncertainty modelling, progresshas been made on the uncoveringof differences between executedflights and CFMU scheduled flights.Characteristics of differencesbetween executed and scheduledflights have been demonstrated tobe time-scale significant: over-deliveries are systematic whentraffic demand decreases in timescales ranging from 5 minutes to 15minutes. However regularities inpatterns characterising uncertaintyare still yet to uncover [INO-04-13to INO-04-13 to INO-04-16].

Implicit Relations in Flight Data


IMPLICIT RELATIONS IN FLIGHT DATA

Claus Peter Gwiggner


1. Abstract

With this thesis a better understanding of why there are differences betweenthe number of aircraft planned to enter fight sectors and the number thatreally entered them is sought. In this report we summarize the analysis of datafrom different flight sectors. As future work, we motivate a generative modelof planning differences in their context of planned traffic.

2. General Background

Airspace is divided into geographical regions, called sectors. For safetyreasons, no more than a certain number of aircraft is allowed to enter certainsectors during one hour. Such numbers are called sector capacities.

Airlines pose a demand to enter sectors before their take-off by submitting aflight plan to a centralized traffic flow management office.

A flight plan is essentially a time stamped list of waypoints. When demand ishigher than capacity either take-off is delayed or aircraft are rerouted. Wespeak of initial demand and regulated demand of a sector.

Although pilots have to follow their flight plans, there are differences betweenthe number of aircraft planned to enter sectors and the number that reallyentered them (the real demand). By consequence, safety is not alwaysguaranteed and available capacity is not always optimally used. We call thesedifferences “planning differences”.

They are consequences of uncertain events like weather conditions, delays,en-air rerouting or more. Such events are not taken into account by thecurrent traffic planning. At a first glance, planning differences occur from thedeviation of a single aircraft from its flight plan. Analyzing this is difficult fordifferent reasons: Some aircraft may recover delays from the start, others may not.

There are dependencies between aircraft: a conflict resolution is madebecause of the presence of other aircraft, delays due to connecting flightsare because other flights arrive too late and so on.

ATM Modelling


Moreover, for an air traffic controller, it is important how many aircraft arriveand not that single aircraft follow precisely their routes. Thus we study theplanning differences of groups of aircraft and not of single aircraft. We hopeto find regularities in data from such groups.

In the following sections we motivate our approach and summarize theresults of a data analysis.

3. Approach

We analyze past flight data in order to better understand planning differences.This task is poorly formalized but oriented by the three axes:

Relations in Time.

Relations in Space. Relations in Scale.

In the first one, we analyze data from a single sector and from interactions oftwo adjacent sectors. In the second, we compare data from two differentsectors and in the third we investigate the characteristics of planningdifferences with respect to the number of aircraft considered.

What we mean with “relation” can be informally described by '(...) variablesthat tend to (...) occur together in a way not expected on the basis of chancealone', the entry for 'correlation' in the Encyclopedia Britannica [1] but will bemore formalized in the paragraph below.

3.a. Data Description.

We focus on four sectors in the upper Berlin airspace where planningdifferences are reported to occur. The sectors are roughly equal in size. Theaverage traversal time of a sector is ten minutes.

We use regulated demand (number of aircraft planned to enter a sector) andreal demand data (number of aircraft that really entered a sector) for a totalof 141 weekdays in the period June 2003-April 2004 of the four sectorsEDBBUR 1 to 4.

3.b. Theories of Uncertainty.

We consider the data as a finite number of realizations of random variables.For a definition of terms from probability theory and statistics, we refer to [2]and [3] or any introductory book.

As an example, we define

€

REALt1 ; t2S to be number of aircraft entering

sector

€

S between times

€

t1 and

€

t 2 for the real demand. Similarly, we define

€

REGt1 ; t2S for regulated demand of sector

€

S between times

€

t1 and

€

t 2 .

Implicit Relations in Flight Data


Finally, we define

€

DIFFt1 ;2S = REGt1 ; t2

S − REALt1 ; t2S for the planning

differences.

With this abstraction, the data analysis can be characterized as a jointdistribution estimation problem. Statistical learning theory, e.g. [4], serves toinfer characteristics of the distributions. Other theories of uncertainty (e.g.dempster-shafer, fuzzy theory, ...) will not be investigated.

4. Results and Future Work

We have analyzed past flight data oriented by the three axes time, space andscale. Our main results are below. Please see [5], [6] and [7] for detailedinformation.

Distributions of planning differences are bell shaped and zero centeredinvariantly of time, space and number of aircraft considered. This certainlyresults from the high number of different reasons for one aircraft to deviatefrom its flight plan (e.g. delay, imprecise flight plan, weather conditions).

However, these reasons are not all independent from each other andplanning differences are discrete variables. The shape of the distributions ofplanning differences of one sector conditioned on its regulated demand isright skewed.

We currently generalize this observation to interactions with neighboringsectors before interpreting it. Autocorrelation, cross-correlation and linearregression estimations showed that there are no non-trivial lineardependencies in the temporal dimension alone.

We rejected the hypothesis of a same underlying distribution of planningdifferences of two adjacent sectors, found linear decision boundaries betweenthem and also, that non-linear boundaries do not substantially improvepredictive accuracy.

This leads us to the idea to establish a generative model of planningdifferences and regulated demand. With such a model, one could identifyregulated demand in order to minimize expected planning differences.

5. References

[1] Encyclopedia Britannica. On-line Version: http://www.britannica.com.

[2] Probabilites, Analyse des Données et Statistique. G. Saporta. EditionsTechnip, Paris. 1990.

[3] Time Series Analysis. Forecasting and Control, 2nd Edition. G. Box,F.Jenkins. Holden-Day, San Francisco, CA. 1976.

[4] The Elements of Statistical Learning. T. Hastie, R. Tibshirani, J. Friedman.Springer-Verlag. 2001.

ATM Modelling


[5] Some Spatio Temporal Characteristics of the Planning Error in EuropeanATFM . C. Gwiggner, P. Baptiste, V. Duong. Proceedings of the 7thInternational IEEE Conference on Intelligent Transportation Systems.ITSC 2004. IEEE Press. pp. 597-600.

[6] Characteristics in Flight Data - Estimation with Logistic Regression andSupport Vector Machines. C. Gwiggner, G. Lanckriet. InternationalConference on Research in Air Transportation. ICRAT 2004. ZilinaUniversity Press. pp. 335-340.

[7] Implicit Relations between Time Slots, Capacity and Real Demand in ATFM.C. Gwiggner. Proceedings of the 23rd Digital Avionics SystemsConference. DASC 2004. IEEE Press.

Optimal Path Planning for ATFM under Stochastic Weather and Capacity Constraints


OPTIMAL PATH PLANNING FOR AIR TRAFFIC FLOW

MANAGEMENT UNDER STOCHASTIC WEATHER AND

CAPACITY CONSTRAINTS

Alexandre d’Aspremont(1), Devan Sohier(2,3),Arnab Nilim(4), Laurent El Ghaoui(4),

Vu Duong(5)

(1) ORFE departement., Princeton University, USA.

(2) LDCI Laboratory, EPHE, Paris, France.

(3) CReSTIC-LICA, UFR Sciences URCA, Reims, France.

(4) EECS department, University of California at Berkeley, USA.


1. Introduction

In the US, delays in the Air Traffic Management Systems (ATMS) are mostoften caused by weather events, which are stochastic in nature. In practicehowever, for complexity reasons, these stochastic obstacles are dealt withusing overly conservative, deterministic strategies. This means avoiding entirelystorm zones that have a very good chance of vanishing in the near future,which translates into lost airspace capacity and unnecessary delays. In Europe,the situation is slightly different. There, sector capacity constraints rather thanweather are the key limiting factor in ATM. Again, these disturbances arestochastic in nature and there too, overly conservative deterministic routingstrategies result in lost airspace. In this work we try to reduce the amount ofwasted capacity by solving the optimal path planning problem in a frameworkthat models both the weather patterns and the sector capacity constraints bya stationary Markov chain. If we assume that a priority order is given rankingthe various aircraft priorities, our algorithm has a complexity that only growslinearly with the number of aircraft. In simulated examples, as weather andtraffic intensity increase, we show a significant improvement over conservativerouting strategies. We also study large scale numerical performance on asimplified European airspace with a large number of aircraft.

Airspace capacity is very sensitive to the presence of convective weather orsector saturation, both introducing a substantial amount of uncertainty. Air

ATM Modelling


traffic delay due to convective weather has grown rapidly over the last fewyears. According to the FAA 2002, flight delays have increased by more than58 percent since 1995, cancellations by 68 percent. The drastic reduction ofairspace capacity interrupts traffic flows and causes delays that ripple throughthe system. Consequently, weather related delays, which are stochastic innature, were the cause of around 80 of the total delay in most years in theUS since 1995.

Past efforts to address delay in the traffic flow management problem in thedeterministic setting include [4], [1], [2], [3]. In these works, demand andcapacities are treated as deterministic and various traffic flow managementalgorithms are proposed in order to reduce the system delay, given that thesystem capacity is exactly known. However the highly stochastic nature ofweather and capacity constraints, both major causes of delay, are poorlyaddressed by a deterministic framework.

In a previous work [7], we incorporated recourse in the planning processwhere we addressed the single aircraft problem using Markov decisionprocesses (where the weather processes is modelled as a stationary Markovchain) and a dynamic programming algorithm. This approach provided a set ofoptimal decisions for a single aircraft that starts moving towards thedestination along a certain path, with the option of choosing a new pathwhenever new information is obtained, in order to minimize the expecteddelay. Furthermore, it showed how the algorithm could be made optimallyrobust to imprecision in the weather transition matrix. A second paper [5]extended this model for multiple aircraft. The complexity of routing multipleaircraft under convective weather grows exponentially. Assuming that aircraftpriorities were known, the results in [5] provided an approximate solution tothe dynamic routing problem for multiple aircraft that minimizes the expecteddelay of the overall system by accounting for the weather uncertainty.

In this work, we detail a method that accounts for the uncertainty related toboth weather and sector capacity and is thus more adapted to EuropeanATM problems. Again, assuming that aircraft priorities are known, wecompute an approximate solution to the dynamic routing problem formultiple aircraft that minimizes the expected delay of the overall system whilesatisfying both weather and capacity constraints. Relative to the deterministiccase, the complexity of our method grows linearly with the number of aircraftand the number of weather states.

2. Weather Uncertainty Model

Various weather teams (Center for Civil Force Protection (CCFP), IntegratedTerminal Weather System (ITWS) etc ) produce predictions that some zonesin the airspace may be unusable in certain time interval and their predictions



are dynamically updated at every

€

δt = 15min. The later an event is from theprediction time, the more unreliable it becomes. It is reasonable to assumethat we have a deterministic knowledge of the weather in the time interval of0–15 minutes in future.

We can discretize time in

€

1,2,L,T stages according to the weather update.We suppose that the weather is in one of

€

S states

€

{1,L,S} and that itsevolution is driven by a Markov chain with transition matrix

€

P such that

€

Pij isthe probability of the system to transition from state

€

j to state

€

i . As inprevious works, we will see that this Markovian setting is a central element inmaking the path planning optimization problem solvable.

3. Problem Formulation

The model for the airspace is composed of a set nodes

€

x1,L,xN ∈ R2 . Forsimplicity but without real loss of generality, the routes are formed bycomputing the Delaunay triangulation of these nodes. We store the resultinggraph in an (symmetric) adjacency matrix

€

M ∈ RN×N such that

€

Mij > 0 iffnodes

€

i and

€

j are connected.

For each weather pattern, we then compute a new matrix

€

MS where thepresence of a storm blocking node

€

xi is modelled by setting thecorresponding coefficients

€

Mi ,jS to zero for

€

j = 1,L,N .

We also assume that state

€

S has all storms active, while state

€

1 has none. Inour simplified model, storms are modelled by closed triangles in the Delaunaytriangulation.

Figure 1 shows an example of such a network with a few sample paths.

We suppose that

€

m aircraft enter the airspace at time

€

1 at the nodes

€

uk ,1for

€

k = 1,L,m . They are sorted by decreasing priority, i.e. with aircraft

€

uk ,1having routing priority over all the following ones, and so on.

We suppose that the airspace is divided into sectors

€

Si with limited capacity.

Furthermore we impose aircraft separation constraints, meaning that twoaircraft cannot occupy the same node at the same time.

4. Dynamic Programming Solution

Here we detail approximate methods to solve the optimal path planningproblem with weather and traffic uncertainty using dynamic programmingtechniques.

ATM Modelling


Figure 1: A network example.

4.a. Deterministic case

To fix ideas, we begin by presenting the deterministic case. Suppose that wewant to solve the optimal routing problem for aircraft

€

1 going from

€

xorig to

€

xdest . Let

€

di for

€

i = 1,L,N be the optimal distance between node

€

i and

€

xdest , setting

€

di =∞ if no solution has been found yet and

€

ddest = 0 . UsingBellman’s recursive solution we cycle through the nodes to update

€

diaccording to:

€

di+ = min{di , minj:Mij>0( x j − xi + dj} (1)

The algorithm terminates when no value of

€

di changes from one iteration tothe next (the number of iterations necessary is then roughly equal to thediameter of the graph). The optimal path is computed by taking the argumentin equation (1) above.

4.b. Weather uncertainty

We now focus on the more interesting case where the weather is in one of

€

S states

€

{1,L,S} with its evolution driven by a Markov chain with transition



matrix

€

P such that

€

Pij is the probability of the system to transition from state

€

j to state

€

i . Here we let

€

ds ,i for

€

i = 1,L,N and

€

s∈ {1,L,S} , be theoptimal expected distance between node

€

i and

€

xdest if the system is in state

€

s , setting

€

ds ,i =∞ if no solution has been found yet and

€

ds ,dest = 0 for any

€

s∈ {1,L,S} . Using Bellman’s recursion we cycle through the nodes toupdate

€

ds ,i according to:

€

ds ,i+ = min{ds ,i , minj:Mi, jS >0( x j − xi + Pu ,sdu ,j

u=1

S

∑ )}

for each

€

i = 1,L,N and

€

s = 1,L,S . This can be written again as:

€

ds ,i+ = min{ds ,i , minj:Mi, jS >0( x j − xi + EsP (du ,j))} (2)

In [5], the problem is further extended to model the situation where theweather transition matrix

€

P itself is known only to belong to a certain set

€

Π .If we look for a worst case solution, where nature (and the weather) playsagainst us, the robust version of (2) becomes:

€

ds ,i+ = min{ds ,i , minj:Mi, jS >0(max(p∈Π

x j − x i + EsP (du ,j)))}

The algorithm terminates when no value of

€

ds ,i changes from one iterationto the next and the optimal path is computed by taking the argument inequation (2) above. Figure 2 shows different possible routes for an aircraftdepending on weather patterns (three storms in this case).

4.c. Traffic and weather uncertainty

The previous algorithm computes the optimal path for one aircraft,accounting for uncertainty in the weather, while making the solution robust touncertainty in the weather model itself. When multiple aircraft share theairspace, we must account for conflict and sector capacity constraints. In [5], itwas shown that finding a globally optimum routing solution for all the aircraftis a very hard problem, hence we must focus on finding a good heuristicmethod to get an approximate solution. One method discussed in [5]assumes that the relative aircraft priorities are known, the routing problem isthen solved sequentially and aircraft with higher priorities are treated as solidobstacles. This solves the complexity issue but can yield somewhatsuboptimal solutions.

Since the routing algorithm is designed for a single aircraft we can expect thesolution performance to degrade in a multi-aircraft setting just as the optimaldeterministic solution degraded when used in a stochastic world. There in facta curse of optimality, i.e. optimal solutions found in a purely deterministicsetting are very sensitive to a change in input, hence perform poorly in a

ATM Modelling


stochastic environment. Since we can’t include the (stochastic) trajectories ofother aircraft in the dynamic programming algorithm (2) without making thecomplexity grow exponentially, we have to include the possibility of conflict ina more rudimentary way.

Figure 2: Sample paths for a single aircraft..

We suppose that we have solved the path planning problem with weatheruncertainty:

€

ds ,i+ = min{ds ,i , minj:Mi, jS >0( x j − xi + EsP (du ,j))}

where

€

MS and

€

P have been defined in previous section 2. The solutionconsists of a vector

€

ds ,iopt ∈ RSN . Here, to account for the fact that conflictand sector capacity problems can perturb the system, we increase the size ofthe uncertainty set to

€

{1,L,S+1} , where the last state corresponds to onewhere all weather zones are blocked and the optimal path computed in

€

doptis also blocked. To keep the model realistic, we attach a small probability

€

νto this state. According to this expanded uncertainty model we form a new



matrix

€

ˆ P and a new adjacency matrix

€

MS+1 such that

€

Mi ,jS+1 = 0 if

€

Mi ,jS = 0

or if node

€

j is the optimal path from node

€

j in state

€

S . We then solve:

€

ds ,i+ = min{ds ,i , min

j:Mi, jS >0

( x j − xi + Esˆ P (du ,j))} (3)

Intuitively, we can expect the solution of this second program to be morerisk-averse than the first one with respect to scenarios where the second-bestrouting choice at a node is significantly worse than the optimal path. On theother hand, for the same reason, we can expect this solution to be morerobust to rerouting due to traffic or sector capacity. The key advantage of thisheuristic is that, while being more robust to traffic perturbations, it keeps thealgorithm parallel, hence computational cost still grows linearly with thenumber of aircraft.

5. Numerical Resuts

5.a. Basic setup

We first construct a simplified model for the European airspace by using a listof 13000 actual beacons, constructing the route network by Delaunaytriangulation, see Figure 5 for a partial picture.

To reduce the complexity of the flight plan computation and without loss ofgenerality, we only take into account beacons that are likely to be flown over.For a plane flying say from b1 to b2, we only retain beacons in a rectanglecontaining the two beacons, and enough room to let the plane avoidpotential no-go zones (weather or congestion areas).

We then compute the distance from each beacon in the zone to the arrivalbeacon thanks to the algorithm detailed in the previous section (MDP). Wefirst set the stopping criterion, and the penalty applied to a flight goingthrough a no-go edge. The result of this program is, for a given flight, an arrayof (expected) distances (one for each beacon and meteorological condition),and an array of “best neighbors”. These two arrays define a routing policy: ateach step, “if possible, go to the best neighbor”. The stopping criterion wechose here is the absence of shift in the best neighbor array and in thedistance array for a number

€

k of steps.5.b. Simulation

To compare our solution to the deterministic one, we use the quantity:

€

ddet − ddynamicddet − dideal

€

dideal being the distance between the departure and the arrival,

€

ddet thedistance with a deterministic algorithm avoiding the no-go zones entirely and

ATM Modelling


€

ddynamic the distance with the MDP algorithm. Thus, this ratio represents apercentage of the best possible improvement.

We compute the optimal routing schemes using the three algorithms detailedin section IV above (deterministic, weather and congestion). Then, weperform a simulation for each algorithm with a given set of planes (identicalfor the three simulations). A plane is defined by a departure beacon, an arrivalbeacon, a departure time and a speed. The speed is, in our simulation, thesame for all the planes. The three other characteristics are chosen at random.We do not take into account planes whose destinations are in potential no-go zones. A plane can fly through a no-go zone only if it is already in it: a planecannot enter a no-go zone. We route the planes using the correspondingpolicy and the three flied distances are recorded, then compared.

5.c. Congestion

We simulate congestion conditions in the following way. We model sectorsby a square paving, 66 square sectors in our case, all with the same capacitywhich user defined.

Since we do not have a priori probabilities for congestion, we must use theresults of the static algorithm to evaluate the parameters of the Markov chainthat models the transition between congested and non-congested states. Thecongested state corresponds to the situations when at least one zone iscongested, i.e. contains more planes than its nominal capacity.

A new parameter for the MDP appears here: the number of potentiallycongested zones we work with. Taking into account all of them does notwork, for some are congested for a very short amount of time, while othersare congested during almost all the simulation. For the

€

k most congestedzones, we compute the transition probabilities, and the final Markov chaintransition probabilities are defined as the average transition probabilities.

5.d. Parameters

a) Meteorological uncertainties only. The stopping criterion is set at 30iterations of the MDP with no shift in the best neighbors array and a totalchange in the distance array less than 1000km. Choosing more extremevalues (more iterations, or a smaller distance) did not induce any significantshift in the best neighbor array in all the tests we led, hence does not changethe policy.

The penalty applied to a flight going through a no-go edge is set sufficientlylarge to make it unambiguously more efficient to avoid a no-go zone than tocross it, storms are circular in our simulation.



Figure 3: Improvement versus weather.

b) Congestion. We use the same stopping criterion in the congestion casethan in the meteorological case. For simplicity, we model congestion by atwo-state Markov chain. This is of course a strong simplifying assumption ascongestion is definitely not Markovian in practice. Of course, a

€

2 j statesMarkov chain, with

€

j the number of potentially congested areas, would givebetter results but at a far more expensive cost. We only consider the threemost congested zones from the deterministic routing simulations.

5.e. Results and discussion

We first test the dynamic programming routing algorithm with weatheruncertainty only. We simulate 100 trials with 2000 aircraft on the simplifiedEuropean airspace with one storm. Figure 3 shows the average improvementmeasure as a function of the stationary probability

€

p of clear weather. Ofcourse, the improvement is always 100% if

€

p = 1 and 0% if

€

p = 0. Asexpected, accounting for weather uncertainty provides a significantimprovement over deterministic routing, even for relatively low values of p.

In Figure 4, we test the dynamic programming algorithm with trafficuncertainty. We simulate 100 trials on 500 aircraft and 66 sectors. Figure 4shows the average improvement measure as a function of the sector capacity,ranging from 2 to 50 aircraft. We pick a (small) congestion probability

€

ν in(3). For very low sector capacities, the entire airspace is almost immediatelysaturated and the algorithm cannot provide any measurable improvement.

ATM Modelling


Similarly, for very large sector capacities, the sectors are almost neversaturated and accounting for the possibility of conflict actually hurtsperformance. However, over a large range of reasonable sector capacities, thealgorithm provides an average improvement of between 10% and 30% overthe deterministic routing.

Figure 4: Improvement versus sector capacity.

Of course, the algorithm in (3) uses an extremely rough model for congestionand a natural extension of this work would be to make the algorithm robustto model uncertainty as in [6].

5.f. Performance

We wrote a C implementation of this algorithms, parallelized using openMP.We compiled it on Romeo, the Sunfire 6800 computer of the University ofReims (SMP 24 °— UltraSparc III @ 900MHz, 24 Go RAM). On 8 processors,the simulation of 2000 aircraft routing with a 4-state Markov model for theweather on our simplified European airspace takes 47 minutes in average. Forthe congestion algorithm and 500 aircraft, it takes 21 CPU minutes onaverage.

6. Conclusion

We provide a traffic flow management tool that can assist Air TrafficControllers and Airline Dispatchers in managing traffic flow dynamically androuting multiple aircraft under weather and traffic uncertainty (the algorithm



produces an optimal route at each network node and each weather pattern).Our proposed strategies deliver a simpler, and on average better, route for anaircraft whose nominal path is potentially obstructed by weather or sectorcapacity constraints. Moreover, these methods naturally reduce the flow ofaircraft in sectors where storms or congestion are expected. As a result, thealgorithms detailed above provide a traffic flow management scheme thatminimizes the expected delay of the system. Relative to the deterministiccase, the complexity of our method grows linearly with the number of aircraftand the number of weather states, when aircraft priorities are known. Largescale simulations on a simplified European airspace show that our algorithmscales well in practice and provides a significant improvement over purelydeterministic routing techniques.

Figure 5: European airspace model.

7. References

[1] D. Bertsimas and Stock Patterson. The air traffic flow problem with enroutecapacities. Operations Research, 46:406–422, 1998.

[2] B. Lucio. Multilevel approach to atc problems: Online strategy control offlights. International Journal of Systems Science, 21:1515–1527, 1990.

[3] H. Marcia. The potential of network flow models for managing air traffic.Technical report, The MITRE Corporation MP 93W0000058, Virginia,USA, 1995.

ATM Modelling


[4] B. Nicoletti M. Bielli, G. Calicchio and S. Riccardelli. The air traffic flowcontrol problem as an application of network theory. Computation andOperations Research, 9:265–278, 1982.

[5] A. Nilim and L. El Ghaoui. Air traffic control under stochastic environments.American Control Conference 04, 2003.

[6] Arnab Nilim and Laurent El Ghaoui. Robust solutions to markov decisionproblems with uncertain transition matrices. To appear in OperationsResearch, 2005.

[7] A. Nilim, L. El Ghaoui, V. Duong, and M. Hansen. Trajectory-based airtraffic management (tb-atm) under weather uncertainty. In Proc. of theFourth International Air Traffic Management R&D Seminar ATM, SantaFe, New Mexico, USA, 2001

Towards a Next Generation of ATM System, Model Based Conflict Detection and Resolution


TOWARDS A NEXT GENERATION ATM SYSTEM

MODEL BASED CONFLICT DETECTION AND RESOLUTION

Dr John Lygeros

Department of Electrical and Computer Engineering, Univ. of Patras, Greece.

1. Introduction

During 2004 effort concentrated in the areas of model based trajectoryprediction, conflict detection and conflict resolution. The work built on earlierresults on modeling and simulation of ATMprocesses, reported in 2003. Theresults of 2004 can be roughly grouped into three categories:

A Monte-Carlo based study of the effect of wind correlation on theprobability of conflict.

An approach to conflict detection based on particle filters.

A randomized optimization based algorithm for conflict resolution.

The publications produced from this work are listed in the references sectionand cited throughout this document.

2. Background: Modeling and simulation

The main aim of the project is to develop tools to assist Air TrafficControllers (ATC) with the task of maintaining separation between aircraft.We will develop algorithms that analyze a given air traffic situation, predictwhether a safety critical encounter is likely to arise in the near future (e.g.over the next 10-15 minutes), inform the air traffic controller of the potentialproblem and possibly suggest ways of resolving it. The work revolves aroundthe idea of using dynamical models for all these tasks. A key element thatneeds to enter into these models is the uncertainty of the process. Thisuncertainty arises from a number of sources, e.g. the wind and weather, themass of the aircraft which is typically unknown to ATC, etc.

The first step in this direction was the development of a physically motivatedmodel to predict the future trajectories of aircraft. The model, which waspresented in greater detail at the Innovative Research Workshop 2003, hasbeen implemented in an object oriented simulator coded in Java. It allows oneto capture many flights taking place at the same time. With each aircraft weassociate a flight plan (based on data from CFMU), aircraft dynamics (with

ATM Modelling


parameter values obtained from the BADA database) and a flightmanagement system (based on the BADA documentation).

The evolution of flights is also affected by the weather, in particular windspeed. Therefore, the evolutions of different flights are coupled to oneanother through a wind model. We model the wind as the sum of twocomponents: a deterministic component reflecting the nominal value of thewind available to ATC through meteorological predictions and a stochasticcomponent reflecting the difference between the actual wind that the aircraftexperiences from the nominal wind. The values of the deterministiccomponent of the wind are based on the Rapid Update Cycle (RUC) servicedeveloped by NASA. The stochastic component exhibits a fairly complexspatio-temporal correlation structure. The qualitative and quantitativeproperties of this structure were based on data found in the literature, dealingwith the statistics of deviations between actual wind and RUC predictions.

An overview of the model was given in a paper presented at the HybridSystems: Computation and Control conference [1] in March 2004 and a two-partpaper [2, 3], submitted to the IEEE Transactions on Control Systems Technologyjournal in December 2004.

3. Wind correlation and conflict probability

The computation of correlated wind samples is rather costly. This is not aserious problem if the model is to be used off line for the validation of conflictdetection and resolution algorithms. If the model is to be used on line,however, for conflict detection and resolution tasks (e.g. based on Monte-Carlo simulations) simplifications may be necessary.

Motivated by this, we set out to estimate the effect of ignoring windcorrelation on the probability of conflict predicted by the model. Even thoughthere are several conjectures in the literature on the strength of this effect, tothe best of our knowledge there has been no systematic study to quantify it.In our study we considered two aircraft in level flight. The probability ofconflict was computed for encounters with minimum separation zero andvarious crossing angles and times to minimum separation. The computationwas done by Monte-Carlo simulation of the model with and without windcorrelation. The results were compared with the predictions of the standardErzberger-Paielli conflict probe. In all cases the conflict probe predictions werevery close to the Monte-Carlo predictions when wind correlation wasswitched off. There was a substantial difference, however, whenever realisticcorrelation was included, especially for shallow crossing angles (45 or 135degrees). Moreover, our results suggest that instead of ignoring correlation, amuch more accurate approximation can be obtained by assuming that thewind is constant (and correlated among aircraft) throughout the encounter.



Early results in this direction were documented in [1]; a more thorough casestudy can be found in [3].

These observations led us to propose an augmentation to the Erzberger-Paielli conflict probe to include terms to capture correlation in the positionsof different aircraft. The predictions of this modified probe closely match theresults of the Monte-Carlo simulations with realistic correlation.

Current work consists of documenting the results of the study and extendingthem to three dimensional encounters. The work is carried out by onegraduate and one undergraduate student at the University of Patras. Weexpect to document the results in a journal publication by the end of 2005.

4. Particle systems in conflict detection

The realistic, detailed models developed so far in this project are ideal forperforming validation of conflict detection and resolution algorithms onsimulated data. However, their complexity limits their possible use fortrajectory prediction for conflict detection purposes. Even after simplificationslike the one discussed in Section 3, the model is still too complicated(nonlinear, hybrid, stochastic) to deal with analytically. The only feasibleapproach for using it in conflict detection appears to be Monte-Carlosimulation.

Analytical complexity is in principle not a problem with Monte-Carlomethods, since the only thing necessary to apply the method is a simulator ofthe process; as discussed above such a simulator is now available for ourmodel. However, computational complexity is still an issue, especially sincethe results of the conflict detection algorithms are needed in real time.Monte-Carlo methods require numerous simulations (typically a fewthousands for conflict detection purposes) of the particular air traffic situationto be executed on-line. Given the current computer limitations, thecomputation may be impossible to perform in real time, even for moderatelycomplex aircraft models and air traffic situations.

Motivated by this we set out to find a way of accelerating Monte-Carlosimulations for conflict detection.

The key observation in this direction is the fact that simulations performed ata given time instant to estimate conflict probability may also be useful infuture time instants. Roughly, if N simulations (say N = 1000) are used attime t, one can assign weights to these simulations according to how wellthey match the radar observations at time t+1. One can then keep the mostpromising simulations and extend them by one time instant, discard the oneswith low weights and initialize new simulations to replace the discarded ones.The predictions of these simulations will then be weighted by the

ATM Modelling


corresponding weights to produce conflict probability predictions at time t+1.When a new measurement comes in at time t+2 the process is repeated.

Particle systems provide a formal way for carrying out this process. In 2004, aPhD thesis [6] on the use of particle methods for conflict detection purposedwas completed at the University of Cambridge, under the joint supervision ofDr J. Lygeros and Dr J. Maciejowski. Current work concentrates on themathematical formalization of some of the empirical results in the thesis; thistask is carried out by a graduate student in the mathematics department atthe University of Patras who is working part time on the project. The hope isthat these results can subsequently be used to speed up the simulationsnecessary for the randomized approach to conflict resolution presented inSection 5.

5. Randomized conflict resolution

For the problem of conflict resolution we proposed a new approach basedon concepts from randomized optimization. The aim in this case is to select aresolution maneuver that minimizes a certain cost criterion that reflects safetyand efficiency considerations. The optimization is carried out using randomselections. Roughly speaking, a resolution maneuver is selected at random,according to a certain search distribution. The expected cost of the maneuveris then computed using Monte-Carlo simulation. The maneuver is “accepted”with some probability, that depends on how the computed cost compareswith the cost of maneuvers selected earlier. Under certain mild assumptionson the search distribution it can be shown that the distribution of theaccepted maneuvers concentrates around the global optima of the costfunction.

The method was applied to two simplified case studies to illustrate itspotential: the sequencing of arrivals in the TMA and the extended TMA(ETMA). The two case studies (along with an overview of the proposedapproach to conflict resolution) were presented in [4] and [5] respectively.

The main advantages of the proposed approach are that:

It is computationally efficient and amenable to sequential and/or parallelimplementation.

It can accommodate very complex trajectory prediction models.

It can accommodate very complex cost criteria.

Explicit performance guarantees can be provided.

The disadvantage is of course that it is still rather computationally intensive,most likely beyond the capabilities of current computers for on-lineimplementation. Particle methods, discussed in Section 4 in the context ofconflict detection, provide some hope in this respect. A more subtle



disadvantage is that, because the method is randomized, only probabilisticguarantees are possible: one cannot be sure that the resulting maneuver isconflict free, even though the probability of it being conflict free can be madearbitrarily high.

Current work concentrates on alleviating the first drawback by developingsequential, particle based implementations of the basic algorithms. We arealso working on a realistic case study of landing sequencing; the formulation ofthis problem is carried out by an undergraduate student at the university ofPatras.

6. Logistics

In January 2004, a graduate student, Ioannis Fourkiotis, was recruited at theElectrical and Computer Engineering department of the University of Patrasto work on the project. Unfortunately, in March 2004 he decided to abandonhis PhD studies to work in industry. Two more graduate students, IoannisLymperopoulos and Adamandia Diamandopoulou were subsequentlyrecruited at Electrical and Computer Engineering, in July 2004. After an initialburn in period, they officially started working on the project in January 2005.In November 2004 one more graduate student, Aristotelis Klamarias, wasrecruited to work on the project part time; Mr. Klamarias is pursuing a PhDdegree at the Mathematics Department of the University of Patras. Finally,two undergraduate students in Electrical and Computer Engineering, GeorgiosChaloulos and Anastasia Salichou, are working on the project for theirdiploma thesis since March 2004.

The close collaboration between the PI and the Engineering Department ofthe University of Cambridge continued. A number of people were involved inwork that relates to this project from the Cambridge side: one postdoctoralresearcher (Dr. A. Lecchini), one PhD student (Mr. O. Watkins) and aprogrammer (Mr. W. Glover). The Cambridge team is supervised by a seniorfaculty member (Dr. J. Maciejowski, the head of the control group). Theireffort was supported by the European Commission under a separate grant(HYBRIDGE, IST-2002-32460). The collaboration with Cambridgeconcentrated on problems of modeling and simulation for trajectoryprediction (W. Glover), conflict detection (O. Watkins) and conflictresolution (A. Lecchini). During the autumn of 2004, O. Watkins completedhis PhD thesis on problems of probabilistic conflict detection [6] under thejoint supervision of Dr. Lygeros and Dr. Machejowski. The collaboration withthe Cambridge team was maintained through mutual regular visits: J. Lygerosto Cambridge 5-7/4/2004, W. Glover to Patras 19-23/1, 17-28/5, and 15-26/11/2004, O. Watkins to Patras 29/11-3/12/2005, and A. Lecchini to Patras,15-21/2, 24/5-4/6, and 8-19/11/2004.

ATM Modelling


7. References

[1] W. Glover and J. Lygeros. A stochastic hybrid model for air traffic controls imulation. In R. Alur and G. Pappas, editors, Hybrid Systems:Computation and Control, number 2993 in LNCS, pages 372–386.Springer-Verlag, Berlin, 2004.

[2] W. Glover, J. Lygeros, and J.Maciejowski. Modeling and simulation of airtraffic management processes. Part I: A stochastic hybrid model. IEEETransactions on Control Systems Technology, 2005.

[3] W. Glover, J. Lygeros, and J.Maciejowski. Modeling and simulation of airtraffic management processes. Part II: Flight management system controllersand simulation. IEEE Transactions on Control Systems Technology, 2005.

[4] A. Lecchini, W. Glover, J. Lygeros, and J. Maciejowski. Air traffic control inapproach sectors: Simulation Examples and Optimization. In L. Thiele andM. Morari, editors, Hybrid Systems: Computation and Control, LNCS.Springer-Verlag, Berlin, 2005.

[5] A. Lecchini, W. Glover, J. Lygeros, and J. Maciejowski. Air traffic controlwith an expected value criterion. In IFAC World Congress, Prague, CzechRepublic, July 4-8 2005.

[6] O. Watkins. Stochastic Reachability, Conflict Detection and Air TrafficManagement. PhD thesis, University of Cambridge, 2005.

Absorption Areas Utility in ATFM


ABSORPTION AREAS UTILITY IN ATFM

Frédéric Ferchaud(1,2), Cyril Gavoille(2), Mohammed Mosbah(2,3), Vu Duong(1)


(2) Université de Bordeaux, LaBRI, Bordeaux, France.

(3) ENSEIRB, Bordeaux, France.

1. Introduction

Airspace congestion is a critical issue European Air Traffic Management(ATM) has to face. The ATCC, Air Traffic Control Centres, capacities are farexceeded by a constant growth in air traffic demand, resulting in everincreasing flight delays. These time and management costs are such a nuisancefor all airlines and passengers that the European Commission has released aspecial statement acknowledging that current ATFM, Air Traffic FlowManagement [1], system are unable to support high traffic loads.

The CFMU, Central Flow Management Unit [2], is in charge of reducing thesecongestion costs by, among several other strategic or tactical measures,delaying departure slots for the flights involved in overloaded sectors(geographical areas of the airspace). The principal objectives of the CFMUhave always been to protect air traffic services against the over-delivery ofaircraft, while at the same time enabling aircraft operators to carry out theirflight operation with the minimum of disruption.

Ground regulations are ineffective in controlling flows when demand is closeto capacity, and yet cause significant delays and reduce aircraft/airportoperators’ flexibility. More accurate flow control would allow increasedeffective capacity and reduce unnecessary delays, while preserving safety.

Considering that the gap between capacity and traffic has been almost closed,effective capacity now needs to grow at an annual rate consistent withEUROCONTROL traffic forecast. This requires careful medium term capacityplanning:

Tailored to individual conditions;

Including flexibility to cope with unexpected events and forecast errors;

Taking safety and cost-effectiveness objectives into account.

This paper presents an improvement of the second point. For this weintroduce the AA, Absorption Areas, corresponding to a safety margin,

ATM Modelling


i.e. flexibility, in order to cope with unexpected events and forecast errorslike: weather conditions;

technical failure;

passenger wait;

flight plan errors...

If there were not uncertainty in air traffic, such that all flights could beguaranteed to respect all estimated time of arrival at all route segments, onecould foreseen a situation in which schedules and routes could be designedto allow gate-to-gate flight without any conflict, and would not be necessaryfor any active intervention by ATCos.

This ideal situation is not yet reachable today; therefore one major issue withATM, Air Traffic Management, is to deal with uncertainty. Current ATM is acomplex socio technical system organizes to cope with uncertainty:controllers and pilots frequently make important decisions based on uncertainor incomplete information, especially in non nominal and emergency situation.

The culture of eliminating uncertainty therefore seems quite deeply ingrainedin ATM, but may not be sustainable or optimal. New concepts and tools,such as those for planning and conflicts detection, tend to increase theamount of data that can be presented and the degree of reliability placed onpredicted information.

The research problem described in this paper concerns the reduction ofdisturbance caused by uncertainty in slot allocation. The goal of slot allocationis to guarantee that the ATCos, Air Traffic Controllers, workload will not beoverload. Actually, the disturbance in slot allocation problem corresponds toaircraft not taking theirs slots and requesting new ones. AA is intended toabsorb these requests without disturbing the scheduler slots. The issue withAA is that full capacity would not be used in pre-tactical planning. An optionaltrade-off shall be found in order to maximize capacity while minimizingdisturbances.

First, one presents the concept of AA and the experimental results. Then onepresents the first results obtain and the distribution of the AA in the slotallocation.

2. Absorption Areas Concept and ExperimentalValidation

When an aircraft lost its slot, it asks for a new one. This demand can delayedother aircraft because of the ”first planned-first served” principle and so forth.



The idea is that aircraft losing their slot will use absorption areas. Anabsorption area is defined as one or several free slots let intentionally unfilledin the slot allocation in order to not disturb the other aircrafts.

To validate this idea, we have developed a simulator for the ”en-route” traffic.The experimentations show the benefits of absorption areas; one can reducethe delays and increase the average throughput of the airspace [3].

3. First Theoretical Results

3.a. Without Reallocation

Then one developed a theoretical approach in order to find an algorithm forthe slot allocation takes into account the absorption areas. Nowadays, oneknows that all the capacity is not declared, but we do not consider it in theslot allocation. One wants to consider it in order to minimize the loadloss,i.e. unused capacity.

One found the best amount of absorption areas according to the rate ofuncertainty [4]. Figure 1 gives the obtained results. The curves represent thenumber of aircraft taking-off under safety conditions with

€

1000 slots, withabsorption areas (the upper one), and without absorption areas.

The previous results were obtained without considering the time of slotreallocation, i.e. absorption areas distribution. The strategy adopted to findthese results is described as follows:

Let

€

n be the number of slots and

€

z the number of unfilled slots (AA)

One allocated

€

(n− z) slots and the AA are at the end of the regulatedperiod.

If an aircraft cannot take its slot, one allocated it a new one in the AA.

With this strategy, one finds the best amount of absorption areas accordingto the traffic demand and the declared capacity. One used the randomwalk [6][7] to find this expected amount, and the Chernoff’s bounds toobtain results with high probability.

ATM Modelling


Figure 1: Obtained results.

3.b. With Reallocation

So it is considered one can used some lost slots. Some disturbances can bedetected enough time before the take off time and so one can used this slotto another delayed aircraft. And even under this assumption, one find benefitsto use absorption areas [5].

4. Absorption Areas Distribution

Nevertheless the previous strategy cannot be applied because of the time ofreallocation. Indeed, delayed aircraft must wait the time of absorption areas,and if the disturbances occurred at the beginning of the regulated period, thistime is not reasonable. One must find an algorithm taking into account thisconstraint (an aircraft having lost its slot must be reallocated in a time lessthan r). Figure 2 shows the strategy adopted to find the results.



Figure 2: Strategy used to find the amount of Absorption Areas.

On Figure 2 one has in red the unfilled slots (absorption areas), in black theaircraft submit to disturbance and having lost theirs slots. The arrowsrepresent the reallocation. Let

€

r be the maximum number of slots we canwait before reallocate these aircraft, it depends of the controller seniority,weather conditions... One considers

€

r equal to

€

4 . Let

€

di be the number ofslots that the delayed aircraft

€

i waits before being reallocated. On thisexample, one has only

€

d3 which is reallocated under

€

r .

One want to guarantee that all aircraft will be reallocated in a reasonabletime. One considers the distribution of the unfilled slots. A first idea is givenby Figure 3. However, one has a big loadloss, i.e. unused capacity.

Figure 3: Example of distribution.

Figure3 shows an absorption areas distribution which guarantee that alldelayed aircraft will be reallocated without delay. In red one has the unfilledslot, and in white the allocated slots. The arrow represents the reallocationability. It could be a good distribution if the rate of uncertainty was greater orequal than

€

50% .4.a. Model

Figure 4 introduces variables used for next results:

Let

€

n be the number of slots.

Let

€

ri be the maximum of slots one can wait before reallocate anaircrafts during the period

€

i . Let

€

zi be the number of unfilled slots of the period

€

i . Let

€

Xi be the random variable representing the number of aircraft wecould not reallocate during the period

€

i . Let

€

m be the number of period in

€

n slots.

ATM Modelling


Figure 4: Distribution of Absorption Areas.

The used strategy is to allocate

€

ri aircraft in the period

€

i , and then addenough unfilled slot to guarantee that most of delayed aircraft will bereallocate in the unfilled slot. The length of the period

€

i is equal to the sumof

€

ri and

€

zi . On Figure 4 one sees how one passes from the strategy usedto find the amount of unfilled slots needed, see Figure 3, to the distribution ofthese slots.

4.b. Probabilistic Results

€

Pr(Xi = 0) corresponds to the case that there are not more delayed aircraftthat unfilled slot in the

€

i period. It corresponds to the sum of thecombinations of

€

0 to

€

zi aircraft in

€

ri slots. So the probability that

€

Pr(Xi = j) correspond to the combination of

€

j+ z delayed aircraft in

€

ri slots.

One is interested by:

One obtains:



For the moment one is interested by the probability

€

Pr that all delayedaircraft find an unfilled slots in its period. So one studies:

4.c. A Simple Case Showing Absorption Areas Benefits

To show the benefits on the distribution, one studies

€

Pr with the sameprobability of

€

pi on all the period, and a same amount of

€

zi in each period.

Figure 5 was obtained with

€

z = r 2 ; it corresponds to an amount of

€

33% ofabsorption areas. One sees on Figure 5 that more

€

r is great, more efficientthe absorption areas are. But with

€

33% of absorption areas and r equal to

€

80, one can guarantee that one will absorb all delayed aircraft for

€

p greaterthan

€

0.7. One needs to distribute this absorption areas more efficiency, inorder to reduce the loadloss, unused capacity.

ATM Modelling


Figure 5: Probability that all delayed aircraft find a slot in a time less than r. r isgiven in a number of slots, p represents the rate of uncertainty.

5. Distribution of Time Reallocation : Robustness andThroughput

One can not guarantee that all aircraft having missed theirs slot can bereallocated in a time less than

€

ri . But one search to reallocate the mostpossible according to the throughput. For this one studies the distribution ofreallocation time of aircraft having missed theirs slots.

On Figure 6 one has the rate of aircraft according to the time of reallocation.If one want a robustness of

€

99% one shows that almost aircraft will bereallocate in a time less than

€

ri but the expected throughput will be reduced,see Figure 7. Actually, more robust the model is, less is the expectedthroughput.

It is why one searches to deal between the traffic demand, the occupancyrate of the slot and the time of reallocation for aircraft having lost theirs slots.One searches to deal between robustness of the model, expectedthroughput and reallocation time.



Figure 6: Expected reallocation time of delayed aircraft accordingto the robustness of the model.

Figure 7: Expected throughput obtained accordingto the robustness of the model.

ATM Modelling


6. Future Work

One continues the distribution study. Then one will take into account thecase of aircraft missing its slot more than one time. These works will befinished for middle of April 2005.

Then one will search to distribute these results in all the regulated system. Forthis one will use the Graph Theory.

Indeed, one has shown the interest of using absorption areas:

A better management of the sectors’ capacity for the disturbances. An improvement of the occupancy rate of the capacity.

So one is encouraged to developed this issue, taking into account the upperairspace. For this, one needs to consider the space time dependenciesbetween the sectors, and the solution one will use to simplify this study, is touse the graph theory. The known algorithm should simplify the problem.

7. Conferences

This work was the subject of several papers published in internationalconferences:

The 7th International Conference on Intelligent Transport Systems held inWashington DC, USA (2004).

The 23rd International Conference on Digital Avionics Systems held inSalt Lake City, USA (2004).

The 1st International Conference on Research in Air Transportation heldin Zilina, Slovakia (2004).

The 3rd International Conference in Computer Sciences, Research,Innovation and Vision for the Future held in Can Tho, Vietnam (2005).

It will also be presented to the MOHSI20052, Méthodologie et Heurisitquespour l’Optimisation des Systèmes Industriels, workshop hold in Hammamet,Tunisia (2005).

8. Conclusion

The first results obtained shown the benefits of absorption areas to improvethe ATFM. Each new assumption reduces these benefits, because in each caseone increases the loadloss. One wants to find an algorithm, according to theprobability of the uncertainty, which guarantees that one can improve the slotallocation.

The experimental results shown that such algorithm must exist. Although theabsorption



areas are already used with the flexibility, they are not yet considered by theCFMU. This safety margin is given according to the ATCos but not accordingto the uncertainty. It corresponds to a continuous absorption area; each hour,one can add the same number of aircraft.

One wants to distribute these unfilled slots more efficiently in the sectors inorder to minimize the loadloss.

In this paper, one sees that taking into account the reallocation time ofdelayed aircraft could also improve ATFM.

Another interest of absorption areas is that if one finds an algorithm giving agood distribution of unfilled slots, then its implementation will neither changesectors topologies, nor controller’s work nor flight plans submissionprocedure.

9. References

[1] Performance Review Report. edition n°7, Eurocontrol, July 2004.

[2] ATFM Users Manual. edition n°9, Eurocontrol, April 2003.

[3] Ferchaud Frederic. Gestion de flux du trafic aerien: un modele de grapheevolutif. Master Thesis, University of Bordeaux 1, june 2003.

[4] Ferchaud, Duong, Gavoille, Mosbah. A new slot allocation for ATFM. inProceedings of the 7th International Conference on Intelligent TransportSystems. Washington DC, Oct. 2004.

[5] Ferchaud, Duong, Gavoille, Mosbah. Using Absorption Areas to improveATFM. In Proceedings of the 23rd International Conference on DigitalAvionics Systems. Salt Lake City, Oct. 2004.

[6] William Feller. An Introduction to Probability Theory and its Applications.volume I, third edition. John Wiley and sons, Inc., 1968.

[7] William Feller. An Introduction to Probability Theory and its Applications.Volume II, second edition. John Wiley and sons, Inc., november 1971.

Performance of Air Traffic Flow Management


PERFORMANCE OF AIR TRAFFIC FLOW MANAGEMENT

Nabil Belouardy


Ecole nationale supérieure des telecommunications, Paris, France.

1. Purpose

This thesis aims to improve the performance of Air Traffic Flow Management,which is about balancing traffic demand to the available system resources.

We have started the thesis by checking the coherence of the data. In fact weneed to know the way how uncertainty generates congestion in the airspace,so as we can apply adequate changes to the current regulation process, orpropose any other strategy able to guarantee a congestion free traffic in theoperational conditions.

2. Topic

Air Traffic Management (ATM) is made of two parts, airborne part (functionalcapability) and ground part which comprise:

Air Traffic Services (ATS): flight information service, alerting service, airtraffic advisory service and air traffic control service,

Airspace Management

Air Traffic Flow Management (ATFM).

The latter is a service complementary to Air Traffic Control (ATC). It aims toensure a safe and orderly flow by protecting controlled areas from overloadand optimizing the exploitation of the available airspace resources. In theICAO6 European region, ATFM is applied in three levels:

Strategic planning, performed at least seven days before the day ofoperation, concerns research activity, demand forecasting and discussingsolutions.

Pre-tactical planning, applied up to the day before the day of operation,works out the set of constraints that should be satisfied by air traffic. Thisarticle focuses on this part.

Tactical planning which contains all that decisions token during the day ofoperation.

6 International Civil Aviation Organization.

ATM Modelling


ATFM pre-tactical planning is established by the Central Flow ManagementUnit (CFMU) in Brussels, it describes the adopted regulation based mainly onground holding programs, and results in ATFM daily plans that would beupdated and used in slot allocation process during the tactical level.

Regulation principle is the following, as airport and airspace resources arebounded demand is to be balanced to cope with capacity specifications.There are four classes of limitations, the corresponding traffic volumes areliable for ground holding program (ATFM delay):

AD (aerodrome), it may be global or just about departures or arrivals

AZ (aerodrome zone), related to standard instrument departure (SID)and standard terminal arrival routes (STAR).

SP (special point), for some particular VOR, TACAN, Loran-C or NDB.

AS (airspace), related to some areas like no-go zones and some flows.

Capacity values expressing the available resources are function of multiplefactors. For instance, the number of aircraft an airport is able to receive doesdepend on the workforce (present staff) and weather conditions.

The interface with ATFM regulators (CFMU) and national airspace managers,i.e. air management cells, is fulfilled by flow management positions establishedwithin air traffic control units.

When slots are allocated to aircraft for take-off, air traffic flows are assignedthrough airspace sectors. Each control unit (terminal, approach or en-routecontrol unit) has to handle safety, apply separation standards and coordinatewith closer units. This task requires a good sharing of workload. As far asterminal and approach control are concerned, controllers work quiteupstream of the system to maintain anyway their own requirements (despiteuncertainties and last minute events). It is en-route control that suffers fromunforeseeable delays aircraft have.

An en-route control center is handled by a team of controllers that dividesthe concerned airspace to some groupings, or collapsed sectors, according totraffic density; this is known as opening schemes.

Each grouping has a capacity; this value expressed in a number of aircraft perhour is an estimation of the maximum number of flights controller can handleproperly. It depends on its geometry, routes and flow complexity. Somegroupings are declared to have a capacity of 999 (infinity!), which means thesegroupings are naturally (always) protected from overload, the other oneshaving finite capacities like those of the example above need to be protectedfrom congestion.



As far as the regulation process is concerned, for each received flight planCASA7 uses the estimated off-block time to compute take-off time by addingsome taxi duration, then to assess the times over each regulated area crossedby the trajectory. Aircraft has a position in each constraint First-In-First-Outqueue and the allocated slot corresponds to the maximum of affected delay.

Last years, the European en-route airspace suffers from congestion, mainly thecore area. The complexity of the ATM system induces non-linear dynamicsthat give an irregular response to the perturbations of operational aleas.

It’s not fair to accuse the regulation process to be deterministic. Uncertaintyof the real traffic is a pretext able to hide the defects of ATFM system design.

3. Study

Using a local simulator named COSAAC8 and its database, the space-timeproperty of congestion seems to depend crucially on airspace size and timescale:

In a wide area, the absence of global congestion can hide concentration ofthe traffic in a sub-region.

In a wide span time, the absence of global congestion can hide large ratesof simultaneous arrivals.

In the case of global congestion, the situation is the same when finest timescale or airspace size is considered.

The first problem (bug) we found in the data concerns flight levels. As theairspace is vertically organized, a waypoint can refer to several sectorsaccording to the flight level. So we have found in a longitudinal sample area:

Some flights related to the lower airspace are filled in upper airspace.

Others flying at higher flight levels are stored in the lower airspace sectors.

After correcting flight level information, there was still another problem (bug)affecting the reliability of the data?

Some flights mentioned in the final traffic demand have not come at all,maybe cancelled flights.

Some aircraft have crossed the sector indeed (according to current trafficdata) without being mentioned in final or regulated traffic files.

The traffic shape constituted of the other flights presents a negligiblefluctuation of the real traffic around the regulated traffic. This can’t justifythe role of uncertainty in creating congestion.

7 CASA : Computer Assisted Slot Allocation.8 COSAAC : COmmon Software to Assess ATFM Concepts.

ATM Modelling


As the data belongs to London Area Control Centre, we have supposed thatthis abnormal traffic is a consequence of the Jet Stream (strong air currentthat can generate traffic variability). Nevertheless, these flights are not comingfrom the American continent.

We have also verified that there have not been rerouted from adjacent oreven far sectors. The likely hypothesis is that they have been ignored by thesimulators because of syntax errors in flight plans. Indeed, we have found themissing flight plans in the database, containing some features like “no-point”or “no-route”, often outside the area of London.

Thus, we were in the obligation to process the flight plans ourselves. Theremain difficulties came from the non standard form of some input files. Forinstance, the borderline of a sector is a closed curve that is not respected inIsland ACC.

Once the prototype is ready, reliable snapshots of the air traffic have depictedthe following results: The collapsed sectors suffer from congestion, even in the regulated traffic.

The filed configuration of opening scheme complies with traffic variationduring the day (good prediction).

The process of regulation doesn’t take in consideration opening schemerequirements. It performs at the macro level of the centre.

The process of regulation takes in consideration other constraints, such asairport limitations.

With small perturbations, the latter constraints become sources ofuncertainty. In en-route airspace, the sum of uncertainties affecting thecrossing flows can be smoothed. However, in the busiest airports, theeffect still remains.

4. Results

As shown in the previous section, the important results are the following:

Traffic regulation is performed at the macro level of centers. It doesn’t take inconsideration the requirements of opening schemes.

The reason behind this decision is that delay represents the real cost of ATMsystem. Thus, it is used only when it is necessary. In contrast to aerodromeconstraints, en-route airspace resources are flexible since each team ofcontrollers is able to choose its own configuration of opening scheme.Therefore they are urged to use this flexibility in order to absorb the possibleover delivery due to traffic heterogeneities within the center airspace.

This kind of qualitative heuristics needs to be assessed quantitatively. Howmuch more total delay would result if ATFM en-route regulation is performed



at the level of collapsed sectors (opening scheme groupings) ? And is thischange really useful when uncertainties of the real air traffic are considered?These questions are under investigation.

Regulation process and uncertainty have similar effects, they are not correlatedbut regulation is likely to amplify uncertainty rather than compensating it.

In fact, regulation consists of ground holding programs, and aircraft are rarelyon time (50% of the delay aircraft have at take-off is due to airlines).

These independent causes of delay generate over-delivery in en-routeairspace. Taking a margin in capacity with respect to uncertainty means a sub-optimal use of the airspace and more total ATFM delay. The best choice ofthe margin in this trade-off is under investigation.

The equity principle of the First Planned is the First Served is applied at a locallevel, regardless network effects. It has no impact on the traffic load but itincreases the total delay.

Let’s consider two flight plans requesting the same route segment at the samemoment (concurrence). If the first aircraft has priority in the first sector andthe other has priority in the next sector (because of slight difference invelocities), then the first aircraft will be affected by ATFM delay (slot of thesecond sector) and the other too (slot of the first sector). In conclusion, theyregulation allows them to come in the same moment, and in the newt slot.

The adopted equity principle is not a total relation of order between flightplans, so it can be substituted by a collective surplus (like the minimum oftotal delay) implemented by the ISA9 software.

5. Examples

The Central centre of London region is constituted of six elementary sectors.

9 ISA : Innovative Slot Algorithm, an alternative to CASA.

ATM Modelling


Figure 5-1 – Elementary sectors of London area EGTTC ACC10,cross and longitudinal sections.

Notice that they can be organized in two sets of three sectors: Western andEastern, or vertically: Upper, Medium Upper and Medium Lowerairspace.First, only one grouping (EGLMU) is defined and thus contains all thatelementary sectors. When the traffic grows significantly at 05:30, airspace issplit in two groupings (EGLUS for upper and EGLMS for medium flight levels).The latter is also split in two groupings (western EG25LMW and easternEG26LME) during the peak period from 13:00 to 21:00. The decrease oftraffic demand during the night explains the return to one grouping mode.

Figure 5-2 – Opening scheme of London EGTTC.

The next figure shows the regulated traffic corresponding to this region, youcan compare the traffic load to the required capacity. The values are relatedto half an hour periods. Clearly, the regulated traffic doesn’t abide by openingscheme requirements. Congestion is present even during the night period (03o’clock) when the traffic density is small.

10 ACC : Area Control Center.



Figure 5-3 – Traffic load compared to capacity in EGTTC ACC,values each 30 minutes.

Figure 5-4 – Traffic load compared to capacity in all EGTT ACCs.values each 30 minutes.

The European Convergence and Implementation Program assesses EGTTarea can borrow a traffic load of 360 aircraft per hour. This is respected andno regulation is performed, however the non uniform distribution of trafficdensity leads to serious problems in the control workload (groupinglevel).The next figure shows ATFM delay distribution of the flow crossingEGTT. Regulation is due to the constraints of other en-route areas. The

ATM Modelling


average delay per aircraft that is 1min 40s is a minimum of the distribution, infact this is due the high rate (about 90%) of non regulated aircraft

Figure 5-5 – Histogram of ATFM delay.

The histogram below is slight skew because of the ATFM delay. Otherwisethe curve would be more symmetric.

Figure 5-6 – Histogram of total delay at departure:ATFM delay + time-over uncertainty.



Most of aircraft have a delay bounded within the period of ±15 minutes. Theinduced variation on traffic load concerns time scales less than the hour.

6. References

[1] N. Belouardy, "Airspace congestion: pre-tactical measures and operationalevents", ICRAT conference 2004, Zilina 22-24 September, Slovakia.

Quantitavely Estimating Wake Vortex Safety using P2P Model


QUANTITAVELY ESTIMATING WAKE VORTEX SAFETY

USING P2P MODEL

Yue Xie(1), John Shortle(1), Peter Choroba(2)

(1) George Mason University, Fairfax, Virginia, U.S.A


1. Abstract

Wake vortex encounter risk is a major safety issue in aircraft final approachphase. The evaluation of the risk is crucial when the industry takes the effortto reduce aircraft spacing for the purpose of capacity gain. This paperproposes a hybrid analytical method to assess the wake vortex encounter riskduring the final approach phase. The hybrid method uses probabilistic methodto calculate the risk based on the probabilistic distributions of aircraftpositions and vortex characteristics obtained from Monte Carlo simulations.The proposed method will be more computational efficient than puresimulations to evaluate the probability of a rare event, such as a seriousvortex encounter. In addition, it is able to provide more insights about riskdistributions with respect to time and location. The estimation given by themodel is conservative, which provides effective safety margin to the riskevaluation in a highly stochastic environment.

Keywords: wake vortex encounter, risk estimation, probabilistic model,quantitative safety assessment, physics of vortices, Monte Carlo simulation.

2. Introduction

Wake vortices are a natural by-product of aircraft lift. A lighter weight aircraftencountering the wake vortex of a heavier aircraft will suffer a roll upset andmay lose control. An example of a wake vortex encounter is an incident onMarch 1, 1993, Orlando, Florida [1]. At the point of upset, an MD-88 wasabout 2.5 NM (65 seconds) behind a Delta B-757 while the flight path of theMD-88 was slightly below that of the B-757. The glide slope angle of bothairplanes was 3°. Data from the Digital Flight Data Recorder (DFDR) indicatethat at about 110 feet Above Ground Level (AGL), the roll angle of the MD-88suddenly reached 13° right wing down and the ailerons and rudder weredeflected about one-half of full travel, 10° and 23°. The crew regained controland continued to an uneven landing.

ATM Modelling


Research on airplane wake vortex physics, characterization, forecasting andrelated aircraft separation has been ongoing both in Europe and the UnitedStates. With the aid of large, fast computers to use Large EddySimulation (LES) models, researchers have much better understood thephysics of vortices [2].

According to [3], wake vortices left behind the generating aircraft can bedescribed by near field and far field characteristics. In the near field, thedominating physical processes are boundary-layer separation, roll-up of thevortex sheet, initiation of vortex instabilities, etc. After roll-up, the wakegenerally develops into vortices, two coherent counter-rotating horizontalflows with approximately equal strength. The far field is the region where theimpact of the atmosphere on the wake vortices is dominant, resulting incirculation decay and structural changes.

The position and strength of vortices within the far field are highly subject toaircraft characteristics and meteorological conditions. Under certain situations,vortices can remain in flight corridors for long periods of time. To avoid wakevortex encounters, following aircraft must maintain a safe distance from theleading aircraft to ensure adequate time for the vortices to decay. The FederalAviation Administration (FAA) of the USA and the International Civil AviationOrganization (ICAO) have divided airplanes into several weight classes andestablished safe separation in the terminal approach area for each pair ofairplane categories. Air traffic controllers must abide by the wake vortexseparation rules to space approach aircraft under Instrument MeteorologicalConditions (IMC).

These separation standards are believed to be one of major constraints toaviation system capacity [3][4]. Hinton et al. [5] estimate that an averagecapacity gain of about 12% can be expected by reducing wake vortexseparation behind heavy aircraft by approximately 1.3 NM and reducingspacing behind B757 by 1.2 NM. Although Gerz et al. [3] do not believe suchgains are realistic, they do think modified wake vortex separation can lead totactical and strategic improvements. They project that even the increase of afew slots per day at a busy airport would be of great value.

Kos and Blom et al. [6] warn that wake vortex induced risk should be betterunderstood before new ATM concepts, including the reduction of separation,for departure and landing on busy airports are deployed. Based on the riskassessment methodology proposed by Blom et al. [7], Kos and Blom et al. [6]introduced a probabilistic methodology to evaluate wake vortex induced risk.

The aim of this paper is to demonstrate a new method to analyze the risk ofwake vortex encounters during the final approach phase. The paper isorganized as follows. The next section will briefly describe the numericalrepresentation of vortex characteristics and will emphasize a particular vortex



model, the Probabilistic 2-Phased (P2P) model. The third section will discussthe existent method to evaluate wake vortex safety, and present a newanalytical method to assess the probability of a wake vortex encounter. Anillustration of the proposed method follows, and the paper ends with adiscussion about the assumptions of the method and future work.

3. Numerical Description of Wake Vortices and theP2P Model

Because only vortices in the far field are of safety interest to us, we focus onthe numerical description of a vortex that has rolled up. The initial vortexdistance

€

b0 is proportional to the wing span

€

B , and the proportionalconstant is the span-wise load factor,

€

s . That is,

€

b0 = s×B .

The age

€

t t of a vortex is usually normalized to obtain a non-dimensionaltime,

€

t* = t /t 0 , so that results from various flight stages, even ComputationalFluid Dynamics (CFD) simulations can be compared. The reference time

€

t 0 isdefined as the time it takes for a vortex to propagate one initial vortexdistance downward. Mathematically,

€

t 0 = 2π × s2 ×B2 Γ0The actual circulation

€

Γ of a vortex is normalized by a reference circulation,

€

Γ* = Γ Γ0 , where :

€

Γ0 =Mg (ρ × s×B×V)The descent speed is also normalized:

€

ω* =ω ω0 , where the referencedescent speed is

€

ω0 = Γ0 (2π ×b0 ) .Based on the definitions of those basic parameters, various mathematicalmodels have been developed to describe the decay and evolution process ofvortices. Detailed review of common models can be found in [3], [8], [9].

Different with most of the mathematical wake vortex models, P2P,Probabilistic 2-Phase decay Model, proposed by Holzapfel, see [10][11][12], isable to accommodate the stochastic nature of turbulence, complex vortexinstabilities and uncertainties of environmental parameters by giving bounds toa deterministic prediction according to corresponding confidence levels. TheP2P model depicts the evolution process of a wake vortex in two phases, assuggested in LES, Large Eddy Simulation, data. One phase is a diffusion phase,the other is a fast decay phase. The decay parameters are calibrated toempirical LES data.

The P2P model uses the average circulation,

€

Γ5−15 , over vortex circles withradii from 5 to 15 meters. Advantages of using average circulation as a metricto describe vortex strength are low sensitivity to observation angles,

ATM Modelling


automatic compensation of vortex motion, and smoothing of the variance.The values of circulation in the diffusion phase and the fast decay phase aregiven in equation (3) and (5) in [10]. The onset time of fast decay and decayrate depend on meteorological parameters,

€

ε , the eddy dissipation rate,and

€

N , the Brunt-Vaisala frequency.

P2P takes into account that the turbulence deforms and transports in astochastic way and results in considerable variation in strength and positionsby giving upper bound and lower bounds of vortex positions and strength. Toobtain the bounds, two runs of P2P with different decay onset time andkinematic viscosity are conducted; then a constant uncertainty allowance of

€

0.2×Γ0* is added to (subtracted from) the initial circulation, and an

uncertainty allowance of one initial vortex spacing is employed for lateral andvertical positions. An example of circulation prediction with upper/lowerbounds based on the P2P model is shown in Figure 1.

Figure 1 – Prediction of Vortex Circulation with Boundsbased on the P2P model.

From a safety perspective, the upper bounds of vortex strength and positionare of more interest because they are more conservative.

4. Existing Wake Vortex Safety Analysis

Modeling of wake vortex evolution is one step in analyzing encounter risk. Toquantify the various levels of safety related with vortices, researchers mustalso integrate aircraft dynamics, induced roll moment and compensationagainst roll. Further evaluation even should include pilot’s response at themoment of encounter.



NLR proposed a probabilistic methodology and developed a correspondingtoolbox to assess vortex-induced risk [6][7]. The probabilistic methodologyfirst evaluates the wake vortex encounter probability based on the MonteCarlo simulations of aircraft dynamics and wake vortex evolutions; then itidentifies the seriousness of encounters considering vortex strength and thetrailing aircraft’s rolling controllability. WAVIR, WAke Vortex Induced Risk, is thetoolbox developed to integrate the simulation models. To estimate the wakevortex encounter probability, the Probability Density Distributions (PDF) ofaircraft positions and vortex positions and strengths are obtained from theaircraft model and the vortex model. PDF’s are inputted in another MonteCarlo simulation to count encounters.

A statistically significant estimation of such a rare event as wake vortexencounter from pure Monte Carlo simulations may be very computationallyexpensive. We introduce below a mathematical method to evaluate aconservative probability of a vortex encounter based on the P2P model. Themethod provides another perspective to investigate the wake vortex risk bynumerically assessing the encounter risk with vortices at various ages.Furthermore, the conservatism inherent in the probabilistic model and theP2P model will help to provide a safety margin.

5. A Conservative Probabilistic Model

Consider the final approach process described in Figure 2. Aircraft fly adesired 3° glide slope to land and the altitude deviation from the desired pathshrinks while aircraft get close to the runway. Wake vortices generated by aleading aircraft will decay and transport for a certain period depending on thecharacteristics of aircraft and current meteorological conditions. If vorticesstay in the flight corridor for enough time, it is possible that a following aircraftwill encounter the vortices.

To mathematically express the probabilistic event, we identify aircraft as theyarrive in succession by the index

€

I. Thus, if the leading aircraft is

€

I, thefollowing aircraft is

€

I+1. We assume that aircraft

€

I+1 can encounter a wakegenerated by aircraft

€

I, but not by aircraft

€

I−1 or earlier.

For notational purposes,

€

x refers to longitudinal position,

€

t refers to the ageof a vortex,

€

Δt is a small time step, and

€

Δx is a small distance step.

Let

€

wv(y,z ; I,[x,x + Δx],[t, t + Δt]) be the latitudinal and altitudinalpositions of vortex at time

€

[t, t +Δt] and longitudinal position

€

x , the onegenerated by aircraft

€

I. Similarly,

€

acft(y,z ; I+1,[x,x + Δx],[t, t + Δt]) isthe location of aircraft

€

I+1 at time

€

[t, t +Δt] after the previous one and atlongitude

€

[x,x+Δx]. Since

€

Δx and

€

Δt are small, we will use

€

x and

€

t insentences instead of their intervals for simplicity.

ATM Modelling


Figure 2 – The Profile of Final Approach Phase.

Suppose

€

t seconds after aircraft

€

I passed the location

€

x , define the eventthat the following aircraft

€

I+1 passes

€

x at this moment as:

€

Is_ acft(I+1,[x,x+Δx],[t + t +Δt])=1,

then the probability of this event is solvable if the distribution function of theseparation is known. The ranges

€

Δx and

€

Δt are chosen to be very small.Define:

€

Is_wv(I,[x,x+Δx],[t + t +Δt ])=1as the event that the wake generated by aircraft

€

I when it passed

€

x is stillalive at age

€

t , then when a following aircraft flies by

€

x at

€

t , the vortex is stillalive with probability:

€

Prob Is_wv(I,[x,x+Δx],[t + t +Δt]) =1( ) .

We define an encounter as the event that the aircraft position

€

(y,z) is in theestimated bounds of the vortex of age

€

t at

€

x , and the notation is:

€

A(x, t) ≡ acft(y,z ; I+1,[x,x + Δx ],[t,t + Δt])

∈

wv(y,z ; I,[x,x + Δx ],[t,t + Δt])

If there is no following aircraft passing by

€

x at

€

t seconds after the previousone, or the vortex dies out at age

€

t ,

€

Prob(A(x, t)) is

€

0 . Only when there isa following aircraft flying by

€

x at

€

t , AND the vortex is still alivethen,

€

Prob(A(x, t)) can be larger than

€

0 . So let

€

M(I,x, t) be the probabilityof an encounter with the vortex of age

€

t longitude

€

x , its value is:



€

M(I,x, t)= Prob(A(x, t))

= Prob

acft(y,z ; I+1,[x,x + Δx],[t, t + Δt])∈

wv(y,z ; I,[x,x + Δx],[ t,t + Δt]),

Is_ acft(I+1,[x,x + Δx ],[t,t + Δt]) = 1,

Is _wv(I,[x,x + Δx ],[t,t + Δt]) = 1)

Equation 1

We go ahead to formulate the probability of an encounter of aircraft

€

I+1 atlongitude

€

x with a vortex at any age. When vortices at age

€

t0 to

€

tn areconcerned, the probability of an encounter is:

€

Prob(A(x))

= Prob(A(x, t0)) × Δ ′ t Δt

+ Prob(A(x, t1) × Δ ′ t Δt

+ L + Prob(A(x, tn )) × Δ ′ t Δt

where

€

A(x, t0) ,

€

A(x, t1) , … ,

€

A(x, tn ) are mutually exclusive events.

€

Δ ′ t isthe sampling time step, and

€

Prob(A(x, t)) is assumed to be identical in theinterval

€

[t,t + Δ ′ t ].Let

€

M(I,x) = Prob(A(x)) , since

€

M(I,x, t) = Prob(A(x, t)) , then if

€

Δ ′ t and

€

Δt are small enough:

€

M(I,x) = P(A(x, ti ) ×Δ ′ t Δt

i=0

n

∑ ≈ M(I,x, t)dtt= t0

tn

∫Equation 2

Since the strengths of vortices are strongly correlated with vortex ages, theprobability of encountering vortices at various ages can be used to representthe probability of encountering vortices with various strengths. For example,vortices generated by a large aircraft at an age younger than 60 seconds maybe viewed as young vortices, and those older than 60 seconds may be viewedas old. The encounters with young vortices are generally more dangerousthan those with old vortices.

Similarly, let

€

Δ ′ x is the sampling distance step, and when

€

Δ ′ x and

€

Δx aresmall enough, the probability that aircraft

€

I+1 hits vortices with age between

€

[t0, tn ] generated by aircraft

€

I over the final approach path is:

ATM Modelling


€

M(I)= P(A(x0)UA(x1 )U ...UA(xm ))

≤ P(A(xi )) *Δx '

Δxi=0

m

∑

= M(I,xi )Δx '

Δxi=0

m

∑

≈ M(I,x)dxx=X0

XRWY

∫ = U(I)

Equation 3

Equality is achieved in the first line when the events

€

A(xi ) are mutuallyexclusive. Then

€

U(I) is an upper bound of the probability of a followingaircraft encountering vortices at age between

€

t0 and

€

tn generated by itsprevious aircraft over the final approach path. Because a vortex encounter ata certain time and a certain location is a rare event, we expect that the boundwill not be too loose.

To evaluate

€

M(I,x, t) , we notice that:

€

M(I,x, t)

= Prob

acft(y,z | I+1,[x,x + Δx],[ t,t + Δt])∈

wv(y,z | I,[x,x + Δx],[ t,t + Δt])

knowing

Is_ acft(I+1,[x,x + Δx ],[t,t + Δt]) = 1Is_wv(I,[x,x + Δx],[ t,t + Δt]) = 1

× Prob(Is _ acft(I+1,[x,x + Δx],[t, t + Δt]) = 1) × Prob(Is _wv(I,[x,x + Δx ],[t,t + Δt]) = 1)

Equation 4

The first term of the formula is a conditional probability, which is theprobability of a wake vortex encounter given there is a passing aircraft and avortex at age

€

[t,t + Δt] is still alive at the longitude

€

[x,x + Δx]. The



conditional probability is determined by the position of aircraft and that of thevortex, whose probability density distributions are assumed to beindependent. The assumption is reasonable because neither pilot norcontroller can see wake vortices to adjust aircraft position.

6. An Illustration of the Conservative ProbabilisticModel

For the purpose of evaluating wake vortex risk, we must integrate a wakevortex evolution model with an aircraft kinematic model. Because we need toconsider the position distributions of both aircraft and wake vortex resultedfrom the variation of aircraft speeds, weights, wingspans, separations, andmeteorological fluctuation, both the vortex model and aircraft model shouldbe stochastic.

6.a. Aircraft Evolution at Final Approach

Without losing generality, we first study the case that both the leading aircraftand the following one are large aircraft, Boeing 727, Boeing 737 or MD80,etc. The characteristics of the aircraft are list in Table 1. We assume all thethree characteristics follow Normal distributions with means and standarddeviations given in Table 1. The data in Table 1 are used only for theillustration of the model. They are hypothetical and not obtained from thecollection of real data. The mean and the standard deviation are chosen toensure that the maximum landing weight will not be exceeded. For instance,the maximum landing weight for a B727-100 is 64638 kg. The probability of arandom number drawn from Normal(50000, 2000^2) larger than 64638 isaround 10-13. The standard deviation of weight in Table 1 is difficult toestimate because it depends on the load of passengers and luggage.

Weight(kg)

Wingspan(meter)

Speed(mtr/sec.)

Mean 50,000 30 70Std.Dev 2,000 2 4

Table 1 – Hypothetical Aircraft Characteristics.

In this paper, we assume that each aircraft flies a constant, but randomlychosen, speed throughout the final approach. After determining the aircraftspeed, the time it takes an airplane to fly from the final approach fix to acertain point in the approach path can be calculated. Let

€

Tx−X0 be the timeto fly from

€

X0 to

€

x ,

€

Tx−X0 = x v , where

€

v is the flight speed. We assumethat

€

v follows a Normal distribution. Although this technically implies that

€

Tx−X0 mathematically is not normally distributed, since the values of

€

v are far

ATM Modelling


enough from zero,

€

Tx−X0 can be approximated using a Normal distribution.Figure 3 shows the distributions of flight time it takes an airplane to fly to thelocations 4000 meters and 7000 meters away from the final approach fixrespectively, as well as their Normal distribution fit.

The variance of the flight time to a further point from the final approach fix isbigger with no controller’s interference. The increased variance will make thedistribution of separation between two aircraft flatter.

Figure 3 – Flight Time to a Certain Point.

Define

€

T1 ,

€

T2 as the time of aircraft 1 and 2 passing the final approach gate,

€

T1* ,

€

T2* as the time passing a certain point

€

x , the separation at

€

x is

€

T2* − T1* = (T2 − T1) + (Tx−X0,2 − Tx−X0,1 ) .

€

Tx−X0,1 and

€

Tx−X0,2 follow the same

distribution. The expected value

€

E[T2* − T1* ] = E[T2 − T1 ] , and the variance

€

Var[T2* − T1* ] = Var[T2 − T1 ]+ 2×Var[Tx−X0,1 ] .

For instance, if the separation at the final approach fix between two largeaircraft is 64 seconds with a standard deviation of 5 seconds, the averageflight time to the point with longitude of 7000 meters is 100 seconds, and itsstandard deviation is 6 seconds. Then

€

T2* − T1* can be approximated usingthe distribution Normal(64, 5^2+2*6^2). Now the probability of a followingaircraft passing the point at 7000 meters at the moment

€

t seconds after theleading one passes it can be calculated:



€

Prob(acft(I+1,7000, t) = 1)= Prob(T2

* − T1* = t)

≈ Prob(T2* − T1

* < t + Δt) − Prob(T2* − T1

* < t − Δt)where

€

Δt is the simulation time step. If

€

t = 90 and

€

Δt = 0.2 , then:

€

Prob(acft(I+1,7000,90) = 1)≈ Prob(T2

* − T1* < 90.2) − Prob(T2

* − T1* < 89.8)

= 0.0013For the interest of space,

€

Δt and

€

Δx will not be shown from now on, butwhen we talk about t and x, we mean

€

[t,t + Δt] and

€

[x,x + Δx].6.b. Wake Vortex Evolution at Final Approach

We continue to use a homogeneous mix of all large aircraft to illustrate wakevortex evolution at the final approach phase. The major characteristics ofaircraft are shown in Table 1, and the hypothetical values of themeteorological parameters are listed in Table 2. The variables are assumed tofollow Normal distributions. Generally, given a certain level of turbulencerepresented by eddy dissipation rate (or stratification represented by Brunt-Vaisala frequency), the lower the stratification, the longer the lifespan of avortex. The values listed in Table 2 are relatively small with respect tostratification and background turbulence, which result in longer lifespan ofvortices. A more detailed discussion on the effects of these parameters onwake vortices can be found in [13].

Air Density(kg/m3)

Brunt–VaisalaFrequency

EddyDissipation

RateMean 1.0 0.016 3.2*10-5

Std.Dev 0.1 0.002 10-6

Table 2 – Hypothetical Meteorological Parameters.

In this paper, we only consider the evolution of wake vortices on the altitudedimension, and assume that we can ignore the drifting distance of vortices onthe lateral dimension. Based on the values of aircraft characteristics andmeteorological conditions listed in Table 1 and Table 2, we collected data onvortex circulation and descent distance at various ages from Monte Carlosimulations, based on the P2P model. The histograms of example data at age20 seconds, 60 seconds, and 140 seconds are shown in Figure 4 and Figure 5for the upper bounds of circulation and descent distances respectively. Unlessexplicitly stated, the circulations or positions calculated from the P2P modelthat we are using later on in the paper are upper bounds.

ATM Modelling


Figure 4 – Circulations of Vortices at Different Ages.

Figure 4 shows that older vortices have a lower average value and variancefor circulation strength than younger vortices. The phenomenon isreasonable since normalized circulation

€

Γ* (τ ) reduces with the increment ofvortex age

€

τ .

Figure 5 – Descent Distances of Vortices at Different Ages.

The distributions of vortices’ descent distance have larger spread whenvortices get older. For the same reason with the situation of vortex’scirculation described above, the variances of vortex’s descent speed reduce asage increases, but it lead to larger variance of distance because distance is the



integral of speed over time. Figure 5 demonstrates that it is more difficult toexactly predict vertical positions of vortices closer to the runway threshold.

When aircraft fly closer to the runway threshold, the altitude deviations oftheir flight path will be diminishing, as illustrated in Figure 2. Our aircraftdynamics model reflects the change of deviation and the examples of thehistograms of the aircraft altitudes and the upper bounds of wake vortexaltitudes are shown in Figure 6.

In Figure 6, the altitude distribution of a following aircraft at a certain location,e.g. 2000 meters from the final approach fix, is independent of thelongitudinal separation with the leading aircraft. So the distribution of (t = 20seconds) is overlapped with that of (t=100 seconds). Old vortices descendmore than young ones with larger variances, as illustrated by Figure 5. If wedefine a vortex encounter as the event that the following aircraft flies underthe upper bound of vortex altitudes, the point that we can see from Figure 6is that, at a certain location, a following aircraft is more likely to hit youngervortices if it passes by, given the configured meteorological conditions. Forexample, the conditional probability of hitting vortices at age 20 seconds is0.032, while that of hitting vortices at age 100 seconds is 6*10-14. Here weassume that the probability of a trailing aircraft flying under the lower boundof vortices is ignorable.

Figure 6 – Aircraft altitude.

The probability of aircraft passing by when a vortex at age 20 or 100 secondshave to be considered to evaluate the real encounter probability. Because thedesired separation for a large-large flight mix is 2.5 NM, the time separation is

ATM Modelling


around 64 seconds assuming the flight speed is 140 knots. From thediscussion above about the aircraft evolution model

€

Prob(acft(I+1,2000,20) = 1) = 2×10−37

which makes the overall probability of hitting a vortex at age 20 seconds isvery small, actually it is 2x10-37x0.032≈6*10-39.

Because the uncertainty of flight time increases with the distance to therunway threshold decreasing, as shown in Figure 3, the probability

€

Prob(acft(I+1,x, t) = 1) displays an interesting feature with respect to thedistance

€

x and vortex age

€

t , shown in Figure 7. According to the aircraftflight model given above, aircraft enter the final approach with separationstandard, which is around 64 seconds. So the possibilities of aircraft passing ataround 64 seconds are much higher than others. However, with the growthof the variance of flight time when aircraft approach to the runway, thedifferences among aircraft passing probabilities are becoming smaller.

The conditional probabilities of wake vortex encounter vanish along thelongitude due to the reduction of the variance of the flight altitudes. Theconditional encounters are less likely to happen to old wake vortices sincethey are further away from the glide slope than younger vortices. Figure 8displays the examples of the vortex encounter probability conditioning that afollowing aircraft passes and vortex is alive.

Figure 7 – Probability of Aircraft Passing A Location at Different Time.



Figure 8 – Conditional Probability of Vortex Encounter.

The evolution of the probability

€

M(I,x, t) over the longitude

€

x and vortexage

€

t is more complicated, and it combines all the factors we have discussedabove.

Figure 9 shows the curves of

€

M(I,x, t) when

€

t are less than the assumedseparation standard, 64 seconds.

Some of the curves in Figure 9 are not monotonously decreasing, while thecurves of

€

M(I,x, t) are when

€

t is close to 64 seconds, as shown in Figure 10.

The integral of

€

M(I,x, t) over the final approach path and over vortex agesis:

€

U(I) = M(I,x, t)dtdx = 1.32×10−4t= t0

T

∫x=X0

XRWY

∫The value of

€

U(I) means that the probability of the following aircraft

€

I+1hitting vortices generated by aircraft I should not be greater than 1.32*10-4.

ATM Modelling


Figure 9 – Probability of Vortex Encounter at A Location at Different Time.

Figure 10 – Probability of Vortex Encounter at A Location at Any Time.

If we are interested in the probability of hitting vortices at certain range ofage, the value can be calculated by adjudging the range of t in the integral. Forexample, the probability of hitting vortices at ages younger than 50 seconds isabout 3*10-7, and the marginal probabilities over the longitude is shown inFigure 11.



Figure 11 – Probability of An Encounter with Vortices Younger than 50 sec.

Figure 12. Probability of an Encounter with Vortices at Agebetween 50 sec. and 70 sec.

The probability of hitting vortices at ages between 50 seconds and 70seconds are 1.31*10-4, and the marginal probability distribution is shown inFigure 12. It demonstrates that most of the likely encounters are related withvortices at age around the deployed separation. According to Gerz et.al ([3]),about 80 encounters per year on average at London-Heathrow Airport. The

ATM Modelling


amount of annual operations in 2003 is around 460,700 [14], so theestimated probability of a human-sensible encounter is about 1.7*10-4.Although we cannot use the Heathrow data to verify the vortex riskcalculated from the Conservative Probabilistic Model, it shows that ourestimation is located in a reasonably correct region. A detailed andcomprehensive verification needs carefully collected data not only aboutaircraft operations, but the meteorological conditions.

7. Discussion

The Conservative Probabilistic Model proposed in this paper is based on aconservative model of wake vortex decay and transport, the P2P model.However, other numerical models of wake vortex are also applicable and theestimation of vortex risk is still conservative. Conservatism in risk evaluation isessential because the system involves uncertainties that are difficult toaccurately identify or quantify, and a certain safety margin can help to reducethe impact of these unknown uncertainties.

The major concern about the model is that the estimation might be muchlarger than the true risk. This can be seen from Equation 3. When a part or allof events

€

A(xi ) are not mutually exclusive,

€

M(I) might be much less than

€

U(I) . The upper bound would be close to the true risk only if events

€

A(xi )are nearly mutually exclusive. As an alternate approach, Kos et al. [6] choosethe maximum value of

€

Prob(A(x i )) over

€

x as the induced risk. This is anunderestimate of the overall risk, and is accurate when the events

€

A(xi ) arehighly dependent. Using this approach, for the example given in this paper, weget

€

Max(Prob(A(x i ));x i ) = 1.7×10−6 , compared to the upper boundpreviously computed,

€

1.31×10−4 .

One intermediate approach is to divide the approach path into segments ofequal length, and to assume that encounters are dependent within an interval,but mutually exclusive between intervals. If we divide the flight distance into

€

m sections with equal length, and define the

€

i -th section as

€

Xi and

€

Maxi (Prob(A(x j));x j ∈ Xi ) as the maximum probability of an encounterhappening in the section

€

Xi , then the overall probability is:

€

Maxi (Prob(A(x j))i=1

m

∑ ;x j ∈ Xi )

If we choose the length of a section to be 20 meters, the resulting probabilityis about 8.3x10-6. As we choose a larger length of a section, the resultapproaches Kos’s estimation. As we choose a smaller length, the resultapproaches

€

U(I) .



Basically the proposed Conservative Probabilistic Model is a hybrid analyticalmodel. The methodology is different from the pure simulation method givenby [6], but it uses a numerical probabilistic method to calculate the risk whilesimulations are conducted to obtain probability distributions of aircraftpositions and vortex characteristics. Numerical methods have the advantageof computational efficiency over pure simulations in evaluating the probabilityof a rare event, such as a serious wake vortex encounter. Another strength ofthe analytical model is that it provides more insights on risk distributions interms of time and location than pure simulations.

Risk distributions depend highly on meteorological conditions, the aircraftflight model, and aircraft characteristics. Therefore the graphs and numericalresults given in the paper are only notional. But, a calibrated analysis withcarefully collected data can be conducted based on the method illustrated inthe paper.

The physics of wake vortices can be much more complicated than thatillustrated in this paper. For example, vortices may rebound in stronglystratified atmosphere or due to ground effect. In such situations, the analyticalresult of encounter risk can be very different with that given in this paper.

We assume that the drift in the lateral direction is negligible. This assumptionis conservative because aircraft and wakes are always at the same positions inthe lateral dimension, so there is no possibility of avoiding a wake due tolateral drift. Taking into account the lateral drift will reduce the upper boundof estimation, especially when a crosswind is present.

Although the probabilistic model does not give risk evaluation directly relatedwith vortex strength (circulation) but via vortex ages, it is easier tounderstand and use from the perspective of operations. In risk classificationanalysis, risk categories should be based on vortex ages, while circulationinformation will not be lost in that vortex circulation is highly coupled withvortex age.

With the Conservative Probabilistic Model, we can go further to evaluatewake vortex encounter risk of a specific airport under the current operationprocedure or proposed future procedures, such as reduction of separationand/or reduction of separation variance.

Acknowledgement. Yue Xie and John Shortle would like to thank WayneBryant, Wake Program Manager at NASA Langley for support of this workthrough NIA under task order NNL04AA07T. We also sincerely appreciatethe consistent help from Dr. Frank Holzaelpfel (DLR) on the coding processof the P2P model. This paper solely represents the opinions of the authorsand does not reflect the opinion of the United States government.

ATM Modelling


8. References

[1] National Transportation Safety Board, 1994, Safety Issues Related toWake Vortex Encounters during Visual Approach to Landing. SpecialInvestigation Report NTSB/SIR-94/01. Washington DC, USA.

[2] Hallock, J.N., Greene, G.C., and Burnham, D.C., Wake Vortex Research –A Retrospective Look, Air Traffic Control Quarterly, Vol.6, No.3, 1998,pp. 161-178.

[3] Gerz, Thomas, Frank Holzaepfel, and Denis Darracq, 2001, Aircraft WakeVortices, A Position Paper, WakeNet.

[4] Lang, Steven, Anand Mundra, et al., 2003, A Phased Approach to IncreaseAirport Capacity Through Safe Reduction of Existing Wake TurbulenceConstraints. 5th FAA/Eurocontrol R&D Conference, Budapest, Hungary.

[5] Hinton, D.A., J.K. Charnock and D.R. Bagwell, 2000, Design of An AircraftVortex Spacing System for Airport Capacity Improvement, AIAA 2000-0622, 38th Aerospace Sciences Meeting and Exhibit, Reno, Nevada,USA.

[6] Kos, J., H.A.P. Blom, et al., 2000, Probabilistic Wake Vortex InducedAccident Risk Assessment, 3rd FAA/Eurocontrol R&D Conference,Napoli, Italy.

[7] Blom, H.A.P., G.J. Bakker, et al., 1998, Accident risk assessment foradvanced ATM, 2nd FAA/Eurocontrol R&D Conference.

[8] Jackson, Wayne, 2001, Wake Vortex Prediction, An Overview,Transportation Development Center, Transport Canada.

[9] Hallock, J.N., G.C. Greene, and D.C. Burnham, Wake Vortex Research – ARetrospective Look, Air Traffic Control Quarterly, Vol. 6, No. 3, 1998,pp.161-178.

[10] Holzaepfel, Frank, 2003, Probabilistic Two- Phase Wake Vortex Decay andTransport Model, Journal of Aircraft, Vol. 40, No.2, pp. 323-331.

[11] Holzaepfel, Frank, Robert E. Robins, 2004, Probabilistic Two-Phase AircraftWake-Vortex Model: Application and Assessment, Journal of Aircraft,Sep/Oct2004, Vol. 41 Issue 5, pp.1117- 1126.

[12] Frech, M., Frank Holzäpfel, 2002, A Probabilistic Prediction Scheme forWake Vortex Evolution in a Convective Boundary Layer, Air Traffic ControlQuarterly, Vol. 10, No. 1, pp. 23-41.

[13] Holzaepfel, Frank, Thomas Gerz, and Robert Baumann, 2001, TheTurbulent Decay of Trailing Vortex Pairs in Stably Stratified Environments,Aerosp. Sci. Technol. 5 (2001), pp.95–108.

[14] BAA; 2004, BAA Heathrow Sustainability Report 2003/04, BAA.

Column Generation for Dynamic Short Term ATFM


COLUMN GENERATION FOR DYNAMIC SHORT TERM ATFM

Olivier Richard(1), Rémy Fondacci(1), Wojciech Bienia(2), Maurice Queyranne(2)

(1) INRETS-LICIT, Laboratoire d’Ingénierie Circulation Transport, Bron, France.

(2) IMAG, Grenoble, France

1. Abstract

The aim of this paper is to shortly present the needs for dynamic, short termATFM and to point out the modelling of the resulting optimization problemas an integer linear program. Then a solution method based on columngeneration is proposed and the resulting framework is detailed.

The particular pricing sub-problem linked to this technique is exposed and afirst solution method is described. All of the methods presented here have tobe validated by computational experiments with data on the whole Europeansystem and this is the current work of the team.

Keywords : Column generation, dynamic ATFM, short term ATFM,synchronisation actions, shortest path problem.

2. Introduction

Air Traffic Flow Management (ATFM) is a service that aims at smoothing thetraffic flow, in order to protect air traffic controllers from overload and tominimize the penalties due to congestion.

As proved by the observation of the daily course of air traffic managementthese goals are not fully reached due to uncertainty on the air system. Inparticular, ground regulations cause in some cases unnecessary delays, whilemissing some sectors overload.

Air traffic services anticipate these overloads by underestimating the numberof flights they can handle in each sector with a global decrease of air systemcapacity as a result. A way to cope with this known problem (consequence ofuncertainty on air traffic regulation) is to add a new regulation filter betweentactical regulation and air traffic control: this is dynamic and short term ATFM.Airborne flights become then subject to flow management regulations thatare in this case sometimes called “synchronisation actions”.

In this paper we shortly describe the principle of dynamic and short termATFM but the main focus of this work is the solving of the optimizationproblems arising. Indeed, the spatial and temporal nature of the regulationactions (ground and airborne delays, rerouting, speed control), the number of

ATM Modelling


concerned flights and the almost real time context create challengingoptimization problems. In some aspects, these problems are close from well-studied older optimisation problems in the field of air traffic management butthe new context of dynamic ATFM add some particular constraints in term ofmodelling (accurate description of the four dimensions of the system needed)and of solving (scarce computational time).

So here is presented a global approach of the optimization problem whichlinks the determination of the 4-dimensional trajectories of the flights and theallocation of these trajectories in a method based on the column generation.

This work is part of a more global research project carried out by the LICITlaboratory in partnership with EUROCONTROL for designing a filter forshort term ATFM.

In section 3, we describe the general context of this project, the structure ofthe short term ATFM filter and the resulting optimization problem ofconstructing and allocating feasible 4-dimensional trajectories.

In section 4, we outline the bases of the column generation technique anddetail its application to this optimisation problem.

The pricing sub-problem is studied in section 5.The concluding section 6 lists future tasks needed to complete this project.

3. Integer Programming Model for Dynamic ATFM

The purpose of the CFMU, Central Flow Management Unit, is to provide anATFM service to aircraft operators and to air traffic services, especially airtraffic controllers.

The principal objectives are: smoothing of air traffic flow, protection ofcontrollers against overload and minimization of penalties due to congestion.The main used actions are ground delays.

However the uncertainty on air traffic system makes these objectives hard tofulfil, particularly when the decisions are taken a long time before the problemthey are aimed to solve. The EUROCONTROL Performance Review Reportcovering year 2003 states:

Ground regulations are ineffective in controlling flows when demand isclose to capacity, and yet cause significant delays and reduceaircraft/airport operators’ flexibility. More accurate flow control wouldallow increased effective capacity and reduced unnecessary delays, whilepreserving safety.

In order to minimize these delays and to achieve a more accurate flowcontrol, the main principle of a short term ATFM filter is to take dynamicregulation actions just before the saturation occurs, when the incertitude is



greatly reduced, see Figure 3-1. Airborne and ground flights are thenconcerned by dynamic, short term ATFM.

Figure 3-1 – Place of the Short Term ATFM filter within the ATFM process.

Here is the structure of this filter: first, a regulation step is defined and at eachstep an overload prediction on the whole European airspace is made.

The regulation then aims to find a repartition of the traffic flows in order tosolve the overload problems. The possible regulation actions are: airborne and ground delay,

vertical and horizontal rerouting,

speed control.

The resulting optimization problem is to determine for each concerned flighta feasible (within the performances of the aircraft, the navigation rules…) 4-dimensional trajectory in order to avoid any sector overload while minimizingthe cost of the regulation.

This allocation problem can be formulated as an Integer Programming model:shown in Figure 3-1. See also the notations in Figure 3-3.

ATM Modelling


Minimize

€

cf ,rxf ,rr∈Rf

∑

+ cgrgg∈C∑

f∈F∑ objective function subject to:

(1) [capacity constraint]

€

∀s ∈ S,∀p ∈ P, yf ,r ,s ,pxf ,rr∈Rf

∑f∈F∑ ≤ C s,p( )

(2) [unicity of the chosen trajectory]

€

∀f ∈ F, xf ,rr∈Rf

∑ = 1

(3) [propagation of delays to connected flights]

€

∀f ∈ F,∀r ∈ Rf ,∀g ∈ C f , xf ,r af ,r + k( ) − rg ≤ dg(4) [integrity constraint]

€

∀f ∈ F,∀r ∈ Rf , xf ,r ∈ 0,1{ }(5) [positivity of delays]

€

∀g ∈ C, rg ≥ 0

Figure 3-2 – Integer Programming modelling the optimization problemresulting form dynamic ATFM.

The constraints are to respect sector capacities (constraints 1) for each sectorand at each time period, to allocate to each concerned flight one and onlyone trajectory (constraints 2) and to ensure that connected flights don’t takeoff before the arrival of the flight they are connected to, constraints 3. Thefactor

€

k represents the minimal turnaround time between the landing of aflight and the takeoff of a connected flight.

They are also the usual unicity and positivity constraints on the variables,constraints 4 and 5 . The objective function is to minimize the cost of thewhole policy, including costs of airborne and ground delays and extra fuelconsumption. There is one 0-1 decision variable for each feasible 4-Dtrajectory (

€

xf ,r ) of each flight which represents a big amount of decisionvariables: the variable linked to a trajectory takes value 1 if the trajectory ischosen, 0 otherwise.

Moreover in air traffic a temporal network effect exists: delays in the morningpropagate in the rest of the day due to flight connections. Direct flightconnections are directly embedded in the program (with the decision



variables

€

rg ) to better evaluate delay costs. It’s unrealistic to model the chainof connections for the whole day because of the uncertainty on the airsystem, so only the first level of connection is modelled explicitly in the linearprogram. The other connection levels are taken into account into the costfunction by increasing the costs of the delays in the morning compared todelays in the afternoon. The sector capacities extend over several periods(hourly, quarterly…) in order to obtain a smooth solution.

Set and elements.Set of flights:

€

F , a flight:

€

f ∈ F .Set of sectors:

€

S , a flight:

€

s ∈ S .Set of time periods (for capacity):

€

P , a time period:

€

p ∈ P .

Set of 4-D trajectory for a flight

€

f :

€

Rf , a 4-D trajectory:

€

r ∈ Rf .Time horizon:

€

T , a date:

€

t .Subset of

€

Rf for the restricted master problem:

€

′ R f .Connected flight characterization.

Set of connected flights:

€

C .Set of flights connected to

€

f :

€

C f , a flight of this set:

€

g ∈ C f .

Scheduled departure time of flight

€

g :

€

dg .Cost of holding flight

€

g on the ground for one unit of time:

€

cg .Turnaround time at the airports:

€

k .Decision variables.

€

xf ,r = 1 if trajectory

€

r ∈ Rf is chosen for flight

€

f ∈ F , 0 otherwise.

€

rg = delay of the connected flight

€

g .Network characterization.

Capacity of sector

€

s during period

€

p :

€

C s,p( ) .

Trajectories characterization.Needed capacity of the sector/period

€

s,p by flight

€

f , trajectory

€

r :

€

yf ,s ,r ,p .

Arrival time of flight

€

f using trajectory

€

r :

€

af ,r .Global cost of trajectory

€

r :

€

cf ,r .

Figure 3-3 – Notations used in the Integer Programming formulation.

The determination of the feasible trajectories (set

€

Rf ) is part of the problembut the simple goal of determining all these feasible trajectories by considering4-dimensional rerouting, both spatial and temporal, is already almostimpossible to reach in a real time context. The column generation technique

ATM Modelling


that avoids an extensive enumeration of the 4-D trajectories is thenparticularly adapted to this problem.

Next part will be dedicated to a short description of the column generationtechnique and to the description of its application to the optimizationproblem resulting from dynamic ATFM.

4. Column Generation Applied to 4-D TrajectoriesDetermination and Allocation

Column generation is a technique to solve a huge linear program (calledMaster problem). When the variables of a linear program greatly outnumberthe constraints, theory of linear programming states that many variables willhave a zero value in an optimal solution.

The main idea of the column generation technique is to solve a restrictedproblem (called Restricted Master Problem, RMP) with only a small subset ofvariables at first, and then to increase this subset by adding some promisingvariables not yet in the subset.

These promising variables are those with negative reduced cost, which wouldincrease the value of the objective function if put into the solution basis.

A sub-problem (called pricing sub-problem) finds such variables from the dualmultipliers using the cost structure without enumerating explicitly all variables.

The restricted master problem is solved again with an enriched set of decisionvariable and new dual multipliers are computed.

The pricing sub-problem checks if they are some negative reduced costvariables to add to the subset and so on. This is an iterative process thatstops when the sub-problem doesn’t find new variables to add to therestricted problem.

Here the master problem is to find 4-D trajectories for each airborne and on-ground concerned flight in order to avoid any sector overload whileminimizing the cost of the policy (see Figure 3-2) and the restricted masterproblem is the same with only a subset of decisions variables (see Figure 4-1).

Integrity constraints are relaxed. Variables are linked to 4-D trajectories andthe column generation technique avoids the enumeration of all feasible 4-Dtrajectories.



Minimize

€

cf ,rxf ,rr∈ ′ R f

∑

+ cgrgg∈C∑

f∈F∑ subject to:

(1) [capacity constraint]

€

∀s ∈ S,∀p ∈ P, yf ,r ,s ,pxf ,rr∈ ′ R f

∑f∈F∑ ≤ C s,p( )

(2) [unicity of the chosen trajectory]

€

∀f ∈ F, xf ,rr∈ ′ R f

∑ = 1

(3) [propagation of delays to connected flights]

€

∀f ∈ F,∀r ∈ ′ R f ,∀g ∈ C f , xf ,r af ,r + k( ) − rg ≤ dg(4) [positivity of decision variable]

€

∀f ∈ F,∀r ∈ ′ R f , xf ,r ≥ 0(5) [positivity of delays]

€

∀g ∈ C, rg ≥ 0

Figure 4-1 – Restricted Master Problem.

The method consists of column generation embedded within a branch-and-bound framework in order to obtain integer solutions (see Figure 4-2).Moreover, an Air Traffic Simulator stores static and dynamic data on the airsystem in particular to allow computing realistic trajectories for flights.

The first step of the process is the initialization of the subset of variables inthe Restricted Master Problem by using pre-processed 3-D trajectories (usualroutes from CFMU and other remarkable routes) and current trajectories ofconsidered flights.

The Air Traffic Simulator converts these 3-D trajectories into 4-D trajectorieswith real time data to initialize the RMP (step initialization in Figure 4-2). Thecore process of column generation (step column generation in Figure 4-2)consists of solving the RMP to obtain a relaxed solution and duals multipliers.

The pricing sub-problem then generates negative reduced cost variables bysolving dynamic shortest path problems with additional constraints in order toget feasible trajectories on the air network.

Valuable trajectories found here can be stored in a 3-D routes database toinitialize the process at a next regulation step. The time dimension is verydependent on the aircraft type and on the dynamic conditions of the air

ATM Modelling


system so it’s useless to store 4-Dimensional trajectories: they won’t be usedin the same configuration. Variables with negative reduced cost are added tothe restricted problem.

This loop goes on until the sub-problem doesn’t find new variables to add tothe subset. A branch-and-bound process coupled with rounding heuristics,step rounding process in Figure 4-2, creates a branch-and-bound tree to get aninteger solution. Some variables that were neglected may become attractivein a particular node of this tree: the column generation loop is started again ateach node of the branch and bound tree.

The process stops when an integer solution that meets all requirements isfound or when all branches of the branch-and-bound tree have beenexplored. In this last case the solution found is optimal.

Figure 4-2 – Framework of the solving process by column generation.

The optimization process is based on a loop between the Restricted MasterProblem and the pricing sub-problem. One part of this loop is a linearprogram solving which can be performed very efficiently by commercial code.



However the second part of the loop is the solving of a very specific sub-problem that presents some challenges: this is the subject of the next section.

5. Solving the Pricing Sub-problem

The goal of the pricing sub-problem is to generate a set of variables,representing feasible 4-D trajectories, with negative reduced cost.

Moreover, the optimality assessment of the whole process depends on theaccuracy of the proof of non-existence of such variables. This problem is thekey problem of the method. It presents some specific difficulties:

The pricing-sub problem is embedded in the optimization process that is onlya part of the regulation process that contains the detection of potentialoverloads, the determination of the sets of flights to regulate, the computingof the solution and its checking and application to the flights with probably ahuman in the loop. Time devoted to the solving of the sub-problem is thenvery short (a few minutes).

There is a different problem instance for each of hundreds or thousand flightsbeing controlled, each problem being a specific dynamic shortest pathproblem with a NP-hard computational complexity. It allows considering eachflight particularity (load, type of aircraft) into the trajectory computation andenvisioning parallel processing but it’s heavy on computing time.

An accurate 3-D description of each trajectory is needed in order to take intoaccount additional constraints defining feasible trajectory, preventing fromusing classical shortest path techniques.

The framework of our solution is a branch-and-bound algorithm on the set offeasible trajectories (see Figure 5-1).

The principle is to build a search tree. Each node of the tree represents asubset of feasible trajectories included in the parent subset. The branching ismade by choosing at each intersection of air route a direction and a targetflight level.

The exploration of the tree is made with rules coming from labelling shortestpath algorithms, in particular the A* algorithm. To compute a lower bound onthe reduced cost needed by the A* method, pre-processed minimal costs ofone trajectory from each intersection of air route to each airport is stored.

The estimates of the reduced cost of partial trajectories are also used to cutthe search. Moreover, many branches of the tree are pruned thanks to aprioritization of flight levels based on speed and fuel consumption. Thisframework allows for an accurate description of the 4-D trajectories and fortaking into account in the computation the characteristics of each flight.

ATM Modelling


Figure 5-1 – Branch and bound algorithm for the Pricing sub-problem.

6. Conclusion and Further Work

This paper focuses on the presentation of the optimization problem resultingfrom short-term dynamic ATFM, its modelling by linear programming and onthe description of the solving process based on column generation. The finalgoal of this part of the research is to test the whole algorithm on real dataand to obtain good solutions within short computing time. For this, parts ofthe process will have to be refined: the management of the initialization database, the sub-problems solving, rounding heuristics, branching strategies andthe definition of the cost function and of the capacity.

Here is for each of these topics the work to be done:

The initialization data base has itself to be initialized with a set of routes richenough to allow a beginning of the column generation loop. Moreover the “inand out” rules for this database have to be defined with a potential significantimpact on the trajectory generation loop and on the size of memory needed.

The sub-problem solving has to be tested with real data on the air systemwith particular reference to feasibility constraints and to computing time. Thisis the current work.

Rounding heuristics have to be studied and developed with a great potentialof computational time saving.

Cost functions in optimization in the field of air traffic management are arecurring problem: a bibliography on these aspects is needed as well as areflection on the costs of delays on connected flights.

During the departure slot allocation, CFMU performs an implicit smoothing ofthe flow while regulating the traffic based on hourly capacity. Constraintscapacity has to be tuned in order to insure this smoothing.



In summary, the column generation technique allows an accurate, dynamicand reactive modelling of the air system control problem. We hope it willlead to an efficient, real time solution to the problem raised by the short termATFM filter, but only computational experiments can validate this approachand this will be the focus of our further work.

7. References

[1] J.M. van den Akker, K Nachtigall. “Slot Allocation by column generation”.NLR Technical Publication 97286. Submitted to European Journal ofOperational Research. 1997

[2] D. Bertsimas, Sarah Stock. “The air traffic flow Management problem withenroute capacities”. Operations Research, 46:406--422, 1998.

[3] Desrosiers, Lübbecke. “A primer in column generation”. Les cahiers duGERAD G-2004-02, Jan. 2004

[4] Goldberg, Harrelson. “Computing the shortest path: A* search meetsgraph theory”. Technical report Microsoft Research MSR-TR-2004-24,March 2003

[5] Leal de Matos P.A., Powel P.L. “Decision support for flight re-routing inEurope”. Decision Support Systems 34 (2002) 392-412, 2002.

[6] O. Richard, R. Fondacci. “Recherche d’un plus court chemin aveccontraintes sur un réseau dynamique de routes aériennes”. Rapport destage de DEA. DEA Recherche Opérationnelle, Combinatoire etOptimisation Grenoble, INPG Grenoble, June 2003

[7] R. Fondacci, L. Penhouët, G. Thomas and L. Zerrouki. “Filter for short-term control of en-route sector load”. NOAA 4th PCRD Project, 1996.

[8] R. Fondacci. “Short term ATFM filter, an approach to the problem”.EUROCONTROL Innovative Research Activity Report 2003.

[9] Gutman. “Reach-based Routing, A new approach to shortest pathalgorithms optimized for road networks”. In proceedings of the 6thWorkshop on Algorithm Engineering and Experiments, Jan. 2004

Optimal Flight Level Assignment : Introducing Uncertainty


OPTIMAL FLIGHT LEVEL ASSIGNMENT:

INTRODUCING UNCERTAINTY

Sophie Constans, Nour-Eddin El Faouzi, Olivier Goldschmidt, Rémy Fondacci

INRETS-LICIT, Laboratoire d’Ingénierie Circulation Transport, Bron, France.

1. Introduction

Commercial flights connecting two airports usually select their flight level so asto minimize fuel consumption. Because most commercial airliners have similarcharacteristics, traffic is practically split among very few flight levels, leading tohigh conflict risk between aircraft. Conflicts are managed by the air trafficcontrollers as emergency situations and summon up a large part of theirattention. Controllers are responsible for a specific airspace sector in whichthey can treat only a limited number of conflicts simultaneously. If too manyconflicts are feared in a controller’s sector, aircraft entrance in the sector canbe delayed, which potentially leads to airspace saturation. Importance of theairspace saturation problem is going to grow fast; solutions are therefore tobe found to ease traffic flow, lighten the controllers workload and limit delays.

The work presented here addresses the problem of tactically assigning theirflight levels to the aircraft before take-off so as to minimize the total conflictrisk once they are airborne. In other words, this problem consists in a globalflight plan optimization problem where the aircraft are to be distributedamong time and airspace by modifying their requested flight levels only. Foreach flight plan, the assigned flight level should be chosen close to therequested one so that fuel overconsumption and change in arrival time arelimited. Moreover, the results of the optimization process should be robust tothe uncertainty on the actual departure times and on the meteorologicalconditions the aircraft will have to face during their trip.

Addressing this problem first requires the definition of several feasible levelsfor each flight. Then, since a conflict involves at least two aircraft, all the pairsof possibilities have to be tested, to determine whether or not they provokea conflict. If so, a cost, or local conflict indicator, has to be defined for eachconflict to represent its severity or its probability. Determination of these localindicators should be done considering the uncertainty on the actual departuretimes and on the velocities of the aircraft. Then, a global conflict indicator,depending on the local ones, has to be minimized thanks to appropriateoptimization methods.

ATM Modelling


In the next section of this paper, we present the work done on thedeterministic part of this problem. This first includes a description of itsmathematical statement. Then, the local conflict estimation procedure ispresented, together with the choice of trajectory and conflict modeling, andthe assumptions made on the aircraft trajectories. The deterministic flight levelassignment problem is then solved using our algorithms, and some illustrativecases are presented. Then, a section is devoted to statistical uncertaintymodeling in order to achieve real conflict probabilities. Finally, conclusions aredrawn in the last section.

2. Deterministic Flight Level Assignment

2.a. Mathematical Statement

The flight level assignment problem can be formulated as a binarymathematical program where each variable

€

xf ,l ∈ {0,1} represents a feasibleflight/flight level combination, i.e. the assignment of a flight

€

f ∈ F = {1..F}to aflight level

€

l ∈ L(f) .

€

L(f ) contains all the levels selected for

€

f and maydepend on

€

f . It generally contains

€

λf , the requested flight level (RFL) for

€

f .

Optimally assigning flight levels to the aircraft is equivalent to selecting exactlyone of these variables for each flight, and set it to

€

1, while minimizing a globalconflict indicator. In this study, the global conflict indicator is simply defined asthe sum of all the local conflict indicators

€

χ (f ,f ' ,l,l' ) related to all the feasiblepairs of assignments. This optimization problem can be formulated as follows:

€

Min Z = xf ,lf =1l∈L(f )

F

∑ xf ',l'f '= f +1l'∈L(f ' )

F

∑ χ (f ,f ' ,l,l' )

subject to:

€

∀f ∈F, xf ,ll∈L(f )∑ = 1

∀f ∈F,∀l ∈L(f ),xf ,l ∈ {0,1}

Concretely, the local conflict indicators

€

χ have to be determined for eachpair of feasible assignments, and then, the problem can be solved byoptimization techniques. In the case where the considered instance consists inmore than a hundred flights, heuristic procedures have to be envisaged, sincethis assignment problem is an NP-hard one.

2.b. Conflict estimation

A procedure has been developed to determine the conflict indicators for allthe considered pairs of assignments. This procedure uses a simplified version



of the trajectories followed by the aircraft. In particular, each flight is assumedto follow the shortest straight line from origin to destination, thus having anorthodromic trajectory. This assumption seems quite strong, but ouroptimization method remains valid for flights following angled routes. Then,aircraft are supposed to respect their theoretical departure times, velocitiesand climbing and descending rates exactly, according to their aircraft type andto the altitude, with an arbitrary safety margin. All the airports are supposedto be located at level 0. Finally, it is assumed that the aircraft follow the newsemi-circular rule for the choice of their cruise flight level that is mandatorysince application of 2002 RVSM measures.

A conflict happens if the minimum horizontal separation distance,

€

Dh , and

the minimum vertical separation distance,

€

Dv , between two aircraft are notrespected simultaneously. Here, we choose to develop a conflictdetermination procedure based on analytical handling of the relative distancesand on the determination of the instants when each aircraft crosses theintersection points of its flight path with the other flight paths. Indeed, noprotection zone is defined around the aircraft, but rather around theintersection of their tracks. We use the proposed modeling of the trajectoriesand the aforementioned hypotheses. We first verify whether the two currentflight/flight level combinations correspond to aircraft having intersecting tracks,and being airborne simultaneously. If so, the intersection coordinates aredetermined as well as the intersection angle

€

θ . Then, investigations arecarried on only if the aircraft have close altitudes around this point.

In the case of aircraft having flight paths which intersect in one geographicalpoint only, we focus on a zone situated around the intersection point. Thestudy is restricted to this zone and the orthodromic distance is approximatedby the Euclidian one. With this simplification, the minimum distance separatingthe aircraft during their presence around the intersection point can beexpressed as a function of the instants

€

T1 and

€

T2 when they cross theintersection, of

€

θ , the intersection angle, and of their velocities

€

u1 and

€

u2.

From this expression, a condition on

€

ΔT = T1 − T2 can be inferred, implying

that a conflict exists between these aircraft:

€

Mint

D1,2 (t) ≤Dh ⇔ΔT ≤ κ1,2Dhwith:

€

κ1,2 =u12 + u2

2 − 2u1u2 cos(θ )u12u22 sin2 (θ )

(1)

where

€

D1,2 (t) is the Euclidian distance between both aircraft as a function ofthe time

€

t . To simplify the computations, this expression is overestimated

ATM Modelling


with a constant

€

κ , sufficiently large to be suitable for most cases, that is tosay:

€

ΔT ≤ κDh . Unfortunately,

€

κ cannot be valid for all the cases, since theexpression of

€

κ1,2 in ( 1) above tends to infinity when

€

θ tends to zero. The

procedure applied in the case of small values of

€

θ , i.e.

€

θ ≤ θlim for smallvalue

€

θ lim , is detailed in the following paragraph.

In the case of aircraft having flight paths that are at least partially common, theflight paths intersection is not reduced to a single point, but corresponds to asegment of the orthodromy. The relative distance between the aircraft caneasily be determined along the one-dimensional route considered and theexistence of the instants

€

T and

€

T' when the aircraft come too close andwhen they separate can also be determined, together with their values.

Here again, a safety margin can be introduced by acting on these two instants.This procedure is also applied for intersecting flight paths having small valuesof

€

θ , i.e.

€

θ ≤θ lim , the distance separating the aircraft being approximated bythe distance separating one of them, say aircraft 1, and the projection of theother aircraft on the flight path of aircraft 1.

We can note that the number of conflicts is overestimated, as well as theirdurations. We will say that the aircraft are in potential conflict (i.e. actually inconflict or near to). It is a strong approximation of the conflict but, this modelenables to be sure of taking all the critical situations into account, which, evenif they do not strictly correspond to a conflict, increase the controller’sworkload. The indicators used in the following are all approximations of theconflict durations or occurrences; they will be noted

€

′ χ (.) instead of

€

χ(.) .

In the applications proposed herein, the conflict indicator is defined as afunction of the duration of the potential conflict. We propose the followingexpression:

€

χ'(f ,f ' ,l,l' ) = κDh − Tf ,l −Tf ',l' if

€

θ ≥ θlim (2)

€

χ'(f ,f ' ,l,l' ) = 1[D1,2 (t)≤Dh ](t)dtT

T'∫ otherwise.

where

€

1[D1,2 (t)≤Dh ](t) is equal to

€

1 for all

€

t such that

€

D1,2(t) ≤ Dh and

€

0otherwise. In this way,

€

χ'(f ,f ' ,l,l' ) is a decreasing function of the minimumdistance between the aircraft, or an increasing function of the potentialconflict duration.

2.c. Flight level assignment

Our flight level assignment algorithms are based on a modeling of theproblem by incompatibility graphs [2]. In such a graph, vertices representflight/flight level combinations and a weighted edge represents an



incompatibility between two combinations, its weight being set to thepotential conflict cost

€

′ χ provoked by the corresponding combinations. Ouralgorithms are also based on the definition of a weight

€

W(s) for each vertexs, equal to the sum of the weights of all its incident edges.

The methods developed are greedy iterative procedures designed to selectexactly one vertex per flight in this graph, in such a way that the resulting setof assignments induces a global indicator value as small as possible.

A typical iteration of one of the presented procedures removes at least onevertex from the graph. This vertex is either eliminated from the solution, orselected in the solution.

In the first case, all the potential conflicts it is involved in are definitivelyavoided in the solution. Their weights have to be removed from theproblem, which amounts, for each of the neighbors

€

s' of the eliminatedvertex

€

s , to subtracting the weight of the edge modeling the potentialconflict between

€

s and

€

s' from

€

W(s' ) . In the second case, the considered vertex is definitively selected in the

solution. The potential conflicts involving this flight/flight level combinationare then accepted in the final solution and their marks have to bepreserved in the graph; no update is made on the weights of the adjacentvertices.

The first algorithm, noted

€

E1 , is based on the fact that eliminating edges inthe graph amounts to eliminating conflicts in the solution. At each iteration,the heaviest edge remaining in the graph is thus considered and we try toremove it by eliminating one of its adjacent vertices. As a priority, we firstfocus on the heavier of the two vertices. It is eliminated from the solution,except if it represents the last assignment possibility for the flight it is relatedto. In that case, elimination of the edge means removing of the second vertex,with the same constraints. It may happen that none of the vertices can beeliminated. This means that the corresponding potential conflict cannot beavoided, at least considering the decisions taken in the previous iterations.There are as many iterations as potential conflicts in the problem.

The following algorithms are based on the management of the vertices of thegraph and are inspired from classical algorithms used for the maximumindependent set problem [4]. In the first one, noted

€

V1, one aims at firstselecting the combinations provoking low costs in the solution. Concretely,each iteration selects the vertex having the smallest current weight. Then,since a flight level has been chosen for the corresponding flight, all the othervertices relative to it can be removed from the problem, while updating theweights of their neighboring vertices. Here, there are as many iterations asflights in the problem.

ATM Modelling


The second algorithm based on the vertices, noted

€

V2 , aims at postponingthe selections of the assignments, so that, at each iteration, the currentweights of the vertices have values closer to their actual costs in the finalsolution. In fact, this algorithm consists in treating the vertices by order ofdecreasing weight: each iteration considers the heaviest vertex and aims ateliminating it. It is selected only if it represents the last possible assignment forthe flight it is related to, that is to say if all the other possibilities have alreadybeen brushed aside. There are as many iterations as variables in the problem.

A last algorithm has been developed, noted

€

V1 /V2 , which aims at uniting

€

V1and

€

V2 and accelerating

€

V2 . It is built on the structure of

€

V2 , and tries tomake decisions as late as possible, except when some are obvious, i.e. whenthere are variables having zero weight in the current graph; then

€

V1 is applied.

Finally, when several variables have equal weights at a given iteration, wepropose to favor one involving an assignment to the level requested by theoperator for this particular flight. Namely, when the algorithm aims at firstselecting variables (this is the case of

€

V1), it selects a variable giving anassignment as close as possible to a requested flight level (RFL). On theopposite, procedures aiming at eliminating combinations (

€

E1 and

€

V2 ) rejectvariables giving assignments which are far from the RFL.

2.d. Numerical applications

The numerical tests carried out so far give encouraging results. We proposeto illustrate them with a numerical application concerning flight plans filescorresponding to six consecutive traffic days of September 2003 over Europe,provided by Eurocontrol. These flight plans files are treated independently,and each contains between 23,000 and 28,000 flights.

The following parameters have been chosen for the application. Each flight

€

fmay be placed on its requested flight level (RFL), on the level just below andthat just above the RFL, if they are suitable for the aircraft type operating

€

f ;this hypothesis ensures a weak fuel overconsumption.

€

θ lim is set to 5°. The

separation distance

€

Dh is set to 10 NM and the separation altitude

€

Dv to1000 ft.

€

κDh is set to 10 min. in (2). The results for this test are presented inFigure 2-1.

For each considered traffic day, the potential conflict indicators valuesobtained are transferred on the graph for the five solutions consisting eitherin assigning their RFL to each flight (noted RFL on the figure), either inassigning the levels obtained after the optimization process with one of theproposed algorithms. For all these tests, we get a significant enhancement ofthe conflict indicator, especially when using strategies

€

V2 and

€

V1 /V2 .



The following average computational times (with a Pentium IV, 3 GHz, 2 GoRAM computer) are given here. The first part, which consists in determiningthe coefficients of the objective function, requires about 35 min. Then,procedure

€

E1 requires 27 sec ,

€

V1 requires 21 sec ,

€

V2 requires 50 sec and

€

V1 /V2 requires 47 sec .

€

V1 / V2 , designed for accelerating

€

V2 , enhances onlypoorly its computational time. Anyway, both algorithms give the best resultson the proposed instances as well as for all those tested in our study, andhave reasonable computational times.

Conflict indicator

0

100

200

300

400

500

600

700

800

900

4 sept. 5 sept. 6 sept. 7 sept. 8 sept. 9 sept.

Ind

ic. (

*10

6)

RFL E1 V1 V2 V1 / V2

Figure 2-1 – Test Results.

3. Uncertainty Modelling

The efficiency of the proposed modeling framework in practice relies largelyon the uncertainty on the instant a given flight crosses a given geographicalpoint. For this reason, uncertainty analysis and modeling was conducted inorder to obtain real conflict probabilities instead of worst-case conflict costsfor our objective function.

The conflict probability of two stable flights in the general case could beexpressed as a function of deterministic parameters and of the probability

ATM Modelling


density functions of the error on the instants

€

T1 and

€

T2 , when the twoaircraft pass over the crossing point, see equation (1).

We can think of each

€

Ti ,

€

i =1,2 , as being the sum of two parts: a systematicpart that is equal to a deterministic function

€

T0,i , and a stochastic disturbanceterm

€

εi . So that:

€

ΔT = T1 − T2 = T0 ,1 − T0 ,2 + ε1 − ε2where:

€

εi ,

€

i =1,2 are random variables with probability densities

€

fεi ,

€

i =1,2

Hence, determining a conflict probability for these aircraft can be done via theprobability of

€

ΔT to be less than

€

κ1,2Dh , c.f. equation ( 1) that leads to thefollowing expression:

€

P Mint

D1,2 (t) ≤ Dh

= P ΔT ≤ κ1,2Dh( )

= fε1z− y− ΔT0( ) ⋅ fε2

−y( )dy−∞

∞

∫

dz

−κ1,2Dh

κ1,2Dh

∫ (3)

This made it necessary to model both the deterministic term and thedisturbance term of the time a flight passes a given point, c.f. equation (3).More precisely, the study was focused on the crossing gap

€

Cgap modeling. Itis defined as the difference between actual crossing time and predictedcrossing time:

€

Actual crossing time − Predicted crossing time

Crossing gap: Cgap 1 2 4 4 4 4 4 3 4 4 4 4 4 = ϕ X1 , ...,Xp( )

Deterministic term

1 2 4 4 3 4 4 + ω

Pure randomtermω~ fω

1 2 4 4 3 4 4

The modeling process was performed using CFMU11 correlated flight plansdata and the standard corpus of statistical tools.

3.a. Deterministic component analysis

According to the crossing gap

€

Cgap decomposition, the deterministic term isset to be a function of some known characteristics (referred to as covariates)of the flights

€

(X1 , ...,Xp ) . The nature of the relationship between

€

Cgap and

€

(X1 , ...,Xp ) could be linear or non-linear depending on the modelingframework.

11 CFMU : Central Flow Management Unit.



Note that in linear case, the deterministic term is defined in the followingmanner:

€

ϕ X1,...,Xp( ) = β0 + β jX jj=1

p∑ (4)

The linearity requirement is not as restrictive as it first sounds. When suchlinear model is adopted, the relationship is linear-in-parameters and becausethe scales of both

€

Cgap and the covariates can be transformed, a suitablelinear relationship can often be found.

Statistical techniques used were standard linear models (mainly linearregression, analysis of variance and analysis of covariance) and tree-basedmodels that is a non-linear modeling framework.

For linear model, one important task was covariates selection from allavailable flights characteristics. For this purpose, the most significant covariatesare derived from the examination of the conditional distribution of

€

Cgap ,given for each covariate (see figures below). As result, the following covariatesare found to be significant:

ADEP: Airport of Departure TYPE: Aicraft TypeAO: Aircraft Operator FL: Flight Level on PointDIST: Distance to Point DAY: Week Day

Figure 3-1 – Scatterplots n°1 for conditional distribution of CGAP.

ATM Modelling







Using CFMU correlated flight plans data, one week data from 3 to 9September 2003, it was found that the linear model explains almost 43% ofthe uncertainty variance and could be expressed as a function mainly ofdeparture airport, aircraft operator and aircraft type. Distances of crossingpoint from departure airport or cruise flight level have no great impact [3].

As an alternative to the technique described above, tree-based model wasused. Such an approach uses a principle known as binary recursive partitioningto achieve a non-linear regression.

Basically, at each step of the tree-building process, the values of the covariatesare examined for all possible splits of the data to find the split that mosteffectively separates the dependent variables, here crossing gap, intohomogeneous groups [1].

For continuous covariates the splits are defined by single value and anobservation goes into one node if its value is less than the split value.

For categorical covariates, all possible partitions of the levels into two non-overlapping groups are considered, and observations are split based on whichgroup contains their value for the covariate on which the split is based.

Note that each variable entered in the model may potentially be used todefine the splits at each stage of the modeling process.

From the tree diagram of Figure 3-5 below, we can see that the first split wasbased on flight level (FL) and this produced two nodes. The left node wasfurther split based on the value of TYPE, the right node was split based onthe distance to point (DIST), and so on. Finally the terminal nodes areindicated, each with the value of

€

Cgap that would be predicted for an

ATM Modelling


observation that fell into that node. The most significant covariates are thosefound in the linear model study.

Figure 3-5 – Tree-based model description for CGAP modeling.

This model explains almost 53% of the uncertainty variance of the

€

Cgap thatcorresponds to a correlation coefficient between actual value of

€

Cgap and

the fitted of 0,73. A plot of the actual value of

€

Cgap versus the fitted valuefrom the model is produced in Figure 3-6 below showing a good fit: points lienot far from identity line.

3.b. Random component analysis

Statistical procedures discussed previously have focused on modeling thesystematic part of crossing gap using the most significant covariates. Thesecond analysis reported in this section describes the probabilistic modelingframework for random part.

Two complementary tasks were performed. First, statistical model adjustmentof crossing gap distribution was performed. The idea here is to infer thenormal distribution assumption that is of widespread use in literature. Second,statistical distribution fits of residuals of the linear and tree-based modelswere performed.

The empirical distribution, see Figure 3-7 below, of the crossing gap (

€

Cgap )

shows inadequacy of normal assumption: 79 % of observations in the [mean± standard deviation] instead of 68 % expected for normal distribution. So,the assumption is rejected by both Chi-square Goodness of Fit Test andKolmogorov-Smirnov test.



Figure 3-6 – Plot of actual values of CGAP versus fitted values from tree-based model.

0

50

100

150

200

250

-4000to -

3500

-3000to -

2500

-2000to -

1500

-1000to -500

0 to500

1000to

1500

2000to

2500

3000to

3500

CGAP (sec)

Fre

qu

ency

Normal

Figure 3-7 – Empirical distribution of CGAP versus normal distribution.

Other statistical distributions were tested and the most appropriate one is thegamma distribution defined by:

ATM Modelling


€

f x( ) =λr

Γ r( )xr−1e−λx , E X( ) =

λ

r, V X( ) =

λ

r2

Figure 3-8 – Histogram and Gamma distribution – Crossing gap.

Working with residuals or disturbance component, derived from linear modeland from tree-based model, one expects, at least for the former model, thatresiduals may behave like a Gaussian random variable.

Residuals for both linear and tree-based models depicted in the figures belowexhibit a symmetric distribution with normal shape and Kolmogorov-Smirnovgoodness-of-fit test validates normality of residuals distribution.



(a) Linear Model Residuals.

(b) Tree-Based Residuals.

Figure 3-9 – Empirical distribution of residuals versus normal distribution.

ATM Modelling


4. Conclusions

The aim of this work was to propose a method aimed at delivering to aircontrollers a traffic that natively contains few potential conflicts. It is based ona graph approach of the problem of assigning the flight levels to the aircraft soas to minimize conflict risk. The global method is divided into two steps: apotential conflict estimation procedure, aimed at setting the problem instance,and an assignment procedure, based on a heuristic method. Four assignmentalgorithms have been presented, together with some tests. The resultspresented here are quite encouraging, providing an improvement of about50% of the total potential conflict indicator of the solution found. A first workon the definition of conflict probabilities has also been presented. It showsthat the uncertainty on the instant an aircraft crosses a given geographicalpoint is better represented by a gamma probability distribution than by anormal one, contrary to what we could have expected.

Now, this work opens wide perspectives. First, the trajectory model shouldbe enhanced, and should take into account the fact that the flight paths arebroken lines and not straight lines. Besides, the modifications imposed on theflight plans should not only affect the assigned flight level, but also the routeand the departure time assigned to the flights; research about the assignmentheuristics has to be completed to enhance this part of the procedure. Finally,we consider the definition and the taking into account of uncertainties as amajor centre of interest in our problem.

The precision of the data used being insufficient, it will be necessary toconduct this study again using better data, for instance CPR data (radar data)and weather data. Besides, further efforts on modeling frameworks must beconducted exploring the following directions: linear and tree-based modelcombination and mixture of normal distributions.

5. References

[1] Breiman L., J. H. Friedman, R. A. Olshen, and C. J. Stone. Classificationand Regression Trees. Wadsworth, Belmont, CA., 1984.

[2] Constans S., R. Fondacci, AND O. Goldschmidt. A greedy heuristicmethod for tactical flight level assignment. 5th Triennial Symposium onTransportation Analysis, Le Gosier, France, 8p, 2004.

[3] Fayette J.-M. Analyse et modélisation statistique de l’incertitude destrajectoires d’avion. Rapport de DESS, septembre 2004, sous laresponsabilité de R. Fondacci & N.-E. El Faouzi.

[4] Sakai S., M. Togasaki, AND K. Yamakazi. A note on greedy algorithms formaximum weighted independent set problem. Discrete AppliedMathematics, Vol. 126, pp. 313-322, 2002.


InnovativeResearch

Workshop

Innovative Research Workshop


Last December, the 3rd Eurocontrol Innovative Research Workshop has beenheld at the Eurocontrol Experimental Centre Brétigny. More than 150 visitors(industries, universities, software houses, R&D laboratories, SME, ANSP,airport, airlines, air traffic controllers) coming from several transports domainsnot ATM only and from more than 24 different European countries havebeen welcomed for 2 days workshop. Advanced ATM concept, emergingtechnologies applied to the ATM, fundamental research studies, Eurcontrolcoordinated actions and Eurocontrol Innovative research strategy werepresented and discussed.

In parallel to the workshop, an exhibition area was open where 15 externalpartners have presented their innovative research activities focusing onemerging man machine interface mixing animation and sound; and groundairport application.

The key event of this workshop has been the celebration of the firstEurocontrol Joint Research Lab – EJRL - between the French Ecole Pratiquedes Hautes Etudes – EPHE– and the EEC.

EUROCONTROL INNOVATIVE RESEARCH WORKSHOP

During two days, all Eurocontrol Innovative research projects have beenpresented and floor was given to the attendees to comment our work. Also,the Eurocontrol Advisory Board – IRAB – composed of Eurocontrol andnational experts, with a representation of the EC RTD department, mandatedto review the entire Eurocontrol Innovative research work, to assess theCARE INO programme and its continuation, and the overall INO strategyand work program for the coming years. Strong emphasis was also given toshow the alignment of the EEC INO work plan with the ACARE StrategicResearch Agenda 2 (SRA2).



Mrs. Pierre Andribet, EEC core business manager, on behalf of Jean-MarcGarot, EEC director, opened the workshop by stressing the objectives of thisevent which were:

To present status and results of all EEC sponsored Innovative studies.

To obtain feedback and recommendation for future work.

To debate the facilitator role of the EEC concerning Innovative studies inEurope.

To debate possible extension of this annual workshop beyond EECfunded projects.

The first day was focusing on the CARE INO programme and emergingtechnologies applied to ATM. The entire CARE INO programme has beenreviewed. Each project has provided the audience with the results of theirwork and with the perspectives they were proposing for the continuationover 2005 and 2006. In 2004, CARE INO was supporting 6 projects coveringemerging technologies applied to ATM such as 3D, Augmented and Virtualreality voice recognition application, graphical animation and sound for ATMman machine interfaces, and quantum cryptography applied for air groundcommunication security; and advanced concept for airport of the future andinter-modality; and lastly mathematical modeling of safety occurrences basedon neural network concepts. Some of these projects were also presentingprototypes at the exhibition enabling anybody to really assess the progress ofthese projects.

IRAB gave feedback on each projectpresented and also recommendations forthe continuation of CARE INO for 2005and 2006. It was the first time that suchfeedback was given lively to all attendees.This has been very well appreciated andgave more professionalism for the CAREINO programme management.

The afternoon was focusing on presentingthe PhD work starting for 3 years at theINO research lab. Each PhD gave anabstract of his/her tasks for the comingyears. PhD work mainly focus onadvanced visualization techniques (3D,Augmented and Virtual reality) applied forATM and how the human – thecontroller – can play with these types ofnew technologies. Case studies will be



performed in the context of airport control tower. This is one contributionfrom the EEC INO team in line with the ACARE SRA2 in the field ofinnovative visualization techniques and tower less ATM. Other studies arestarting on the extension of the watermarking technology. Other PhD aremore working for advanced concept in the field of small air vehicle that couldbe largely used in the future, or in the field of the airport of the future. Againthis is a strong contribution in line with the ACARE SRA2 innovative researchlike ATM for new vehicles, highly automated ATM and airport of the future.

On the second day, advanced concepts were presented, especially theparadigm SHIFT project which has developed innovative concept. Thisinnovative concept impacts the airspace typology (highways, tube), theworking methods, the collaboration amongst all ATM actors being on theground at the airport level or in the air, social and legal issues in terms of newcontractual agreement. It has been also recognized that there were some linkswith the MUAC MANTAS project which was also presented from MANTASproject team at the exhibition. This work is also complemented with a set ofPhD studies working on the automation concepts. One is especially focusingon developing some of the ERASMUS concepts in the field of subliminalcontrol and highly automated ATM part of the ACARE SRA2 ATM R&Dprogramme. One of the outputs of the Paradigm SHIFT project presentedwas a complete research agenda that will from now, be the foundations ofthe entire EEC INO advanced concept work plan. This research agendashows the various issues that need to be addressed when defining andvalidating a new ATM concept..



The afternoon of the second day was dedicated to the modeling activities,analytical, uncertainty and soft computing, ongoing at the Innovative researchlab. Then, before closing the 3rd workshop, the innovative research labstrategy for the coming years – 2005 to 2009 - was presented, reviewed andassessed by the IRAB who finally endorsed our strategy for the coming years.

The overall feedback of these 2 days was very positive and we can summarizeit by re-using Jan Van Doorn’s words “I spent 2 refreshing days at the EEC bylistening to new fresh ideas and by having the chance to see and touch innovativeproducts when visiting the exhibition”. We had also very good input from theattendees which will be considered when developing the INO workprogramme. Some of the major recommendations given were to extend thisworkshop to external partners, and also to think about organizing summeruniversities for young ATM European researchers.

What a challenge for us to make next Eurocontrol Innovative Researchworkshop in 2005 better than in 2004 !

FIRST EUROCONTROL JOINT RESEARCH LAB

On Thursday evening, the EEC and the French Ecole Pratique des HautesEtudes – EPHE Sorbonne – celebrated the creation of the first EurocontrolJoint Research Lab. We were glad to see Mss Courtel, President of theFrench Ecole Pratique des Hautes Etudes – EPHE Sorbonne – and Mrs.Andribet, EEC Business Manager, signing the convention. This was the resultof several months of negotiation between the 2 partners.



We had also the pleasure and the great honor to see Mss Courtel, Presidentof the French Ecole Pratique des Hautes Etudes Sorbonne and Mrs. Andribet,EEC Core Business Manager, signing and, thus, celebrating the firstEurocontrol Joint Research Lab.

This first Eurocontrol Joint Research Lab also called in French Unite MixteEuropeenne de Recherche, the first attempt to create a network of Europeanuniversities to perform backbone thinking for ATM, will be specializing in theModeling of ATM System in which human cognition will be modeled andanalyzed from a complex system modeling approach. Note that ATM SystemModeling has been identified within ACARE SRA 2 as an importantprerequisite for future improvements in ATM. This new Joint Research Lab islocated at the EEC, and baptized Laboratory for Complex System Modeling &Cognition (CSMC). As part of the agreement, five senior research scientistsand professors from EPHE are partially seconded to the laboratory toconduct research and to supervise PhD students’ work, and EEC staffs willgradually increase their teaching activities at EPHE.

This joint research lab will enable the sharing of expertise, knowledge andexperimentation means from both sides. Today, limited to 2 initial partners,this joint research lab will seek to get other European partners.

Also, the EEC INO lab will develop this concept of joint research lab onother ATM domain such as, for example, 3D, Augmented and virtual realitytechnologies, or airport of the future.



EUROCONTROL

INNOVATIVE

RESEARCH

WORKSHOP

EXHIBITION

In parallel to the workshop, 15 external partners have presented theirinnovative research activities focusing on emerging man machine interfacemixing animation and sound; and ground airport application.

This was a unique opportunity to see,l i s ten and touch advancedtechnologies and products that couldbe in the future largely deployed inthe ATM world.

A serie of innovative man machineinterface combining animation as usedin the game industry, with soundinteraction were presented fromIntuilab and Intactil design with thekind participation from AppleCompany. This raised a great interestfrom the visitors who saw and touchfor the first time what game industrycould bring to our ATM business.These projects are fully supported bythe CARE INO programme. Decisionhas been made to investigate furtherthe use of these technologies for theATM.

3D, augmented and virtual reality based application were also demonstratedfrom pure in lab work (ARMIN Paris university), and for the industry world(BARCO-Orthogon) which tend to show that these technologies are wellmastered now, and would need a work on their application for the ATM. It isthe rational why EEC INO team is developing projects to assess the use ofthese emerging technologies within the ATM: see tower less projects.



Other man machine interfaces (VertiDigi and Vigiestrip) were presented bythe French CENA which demonstrates real added value of using animatedman machine interface for the ATM.

Voice recognition application was also presented not only to show that thistechnology works, but to propose how it can be used for ATM by ThalesTRT and IRIT. Any visitor was able to speak, to listen and to see from bothground and air side. Other applications focusing on airport groundmanagement were also demonstrated like 2 EC projects: AVITRACK fromSILOGIC and AIRNET from M3SYSTEMS. These types of applications, andthe domain itself, were quite new for the EEC INO team but it shown thatfrom both a technical and business point of views, there were some strongsynergies to develop in the future. Also, this was a good opportunity todevelop INO team involvement in the airport domain bearing in mind thatINO is starting several projects focusing on the airport of the future.

We had also the great pleasure to welcome MUAC people who presentedthe MANTAS project. This was a great opportunity to exchange on advancedATM concept between the MANTAS and Paradigm SHIFT teams. We dohope these first contacts will be developed in the future.

Considering the success of this first exhibition organized within the workshop,and considering the amount of fruitful exchanges between the people whoattended the event, we plan to develop this activity for the next workshops.


CooperativeActions of R&D

in Eurocontrol

CARE INO

CARE INO


CARE stands for Co-operativeActions of R&D in Eurocontrol. Ithas been set-up in 2000 by theEurocontrol Agency to define co-operative actions which addressR&D issues of high priority. Theapproach to the research is co-operative, collaborative; i.e. it fostersmotivation and exchange of ideas,b r ing s toge ther d i f f e ren tapproaches, cultures, competencies,forges common views and solutions.

CARE is structured around several“actions” like CARE ASAS, CAREINTEGRA, CARE AIRPORT, CARETP and CARE Innovative; and eachaction is materialised by a set ofprojects. Up to the end of 2003,the programme was fully managedby Eurocontrol HQ. But, early 2004,the decision was made to put themanagement of the CARE INOaction at the Experimental centre,under the responsibility of the EECINO RA.

The CARE INO in i t iat iveconcentrates on co-operative workaimed at Innovation in ATMResearch. The spirit is to open thefloor for external bodies such asUniversities, R&D centres, small,medium and large industries topropose projects aimed atdevelopping any innovative idea, beit new or emerging technologiesapplied to ATM or new ATMconcepts or a combinaison of both.The objective is also to re-inforcecooperat ion within Europe,attempting to avoid duplication ofwork and more efficient use of the

existing resources within eachorganisations. The projects run bythe organisations are inter-related.Calls for proposal are issued to findorganisations interested in this typeof work and CARE INO is providingwhole or part of the funding todevelop the new ideas.

Generally speaking, one CARE callfor proposal is run over 3 years.The first year is dedicated torefining innovative idea that hasbeen selected by the CAREprogramme, whereas the 2following years are used to furtherdevelop the innovative idea.Therefore, in the first year, severalprojects are running in parallel. Atthe end of the first year, a review ofeach project is performed to decidewhether the innovative aspect isstrong enough or not, and then todecide on the continuation (or not)of each project.

In October 2003, following thesuccess and promising results of theCARE-I INO initiative, a second callfor proposals for innovative ideasfocusing on the theme of Innovativeand Advanced Technologies for theATC/ATM of the future wasinitiated. Further information on thecall for proposal was discussed atthe yearly EEC Innovative ResearchWorkshop on December 2003. 50proposals were submitted beforethe deadline of 31st of January2004. The EUROCONTROLInnovative Research AdvisoryBoard, IRAB, composed ofEurocontrol, EC and FAA experts,

CARE INO


who is in charge of steering theentire EEC INO activities, met onFebruary and advised to select 6out of the 50 proposals. The 6selected proposal covering severalresearch threads are:

Air Ground Telecommunication,AGT, project with ENST (EcoleNationale Superieure desTelecommunications Paris –France) - aiming at evaluatingAir Ground telecommunicationsecurity and its enhancement byusing Quantum Cryptographytechnology.

Adapted observation foractivities of an airport – withARMINES (Ecole des MinesParis – France) and Readymade- exploring the provision andthe sharing of the vision of itsworking environment to anyprofessional branch of theairports actors. This project isbased on augmented reality,mobile and wireless applications,

Safety of Controller PilotdialoguE - SCOPE project withThales Research & TechnologyFrance, IntuiLab and IRIT –exploring the use of voicerecognition systems as a thirdpart in the communication,captur ing parts of thecommunication, and using theresulting information to improvethe efficiency and safety.

ANIMS - with IntuiLab andIntactile design (Toulouse –France) - aiming at improvingefficiency and safety of ATM

users interfaces through visualanimation and sound. Thisproject explores the potentialfor ATM software of the design-centred methods used inindustr ies from dif ferenthorizons. Focusing on studyingthe benefits and conditions ofuse of two related design-intensive interface technologies:animation and sound.

N e u r a l n e t w o r k - b a s e drecognition and diagnosis ofsafety critical events - with theNLR and SNN University ofNijmegen (Netherlands) -aiming at investigating thefeasibility of a neural network-based system for automaticrecognition and diagnosis of nonnominal events in ATM. Suchsystem is intended to furtherenhance safety in ATM.

Airport of the future or centrallink of inter-modal transports -with M3 SYSTEMS, ENAC,LEEA and ANA (Portugal) -exploring the possibility that thetransport modes could becollaborative instead of onlycompetitive, and exploring thetransport inter-modality as away to tackle what could be theATM/ATC. This is an attempt toenvision the airport of thefuture.

These projects have been reviewedduring the 3rd EurocontrolInnovative Research workshop, heldin December 2004 at the EEC.Results and perspectives were

CARE INO


analysed by the IRAB. Decision hasbeen made to select 3 projects forcontinuation in 2005 and 2006:

Thales Research & TechnologyFrance in collaboration withIntuiLab and IRIT ( Paris,Toulouse – France) with SCOPEproject;

IntuiLab in collaboration withIntactile design (Toulouse –France) with ANIMS project

M3 SYSTEMS – ENAC (France)in collaboration with LEEA–ANA (Portugal) with the airportof the future project (central linkof inter-modal transports?).

During the same year, being part ofthe CARE INO objectives,dessimenassion of the results ofCREA! project, one of the firstCARE INO projects , wasperformed. CREA! is an inter-disciplinary approach to design thatencourages intersections betweenart, design and technologyborrowing from each discipline

practices, methods and experiencefor the definition of innovativeconcepts. The application of theapproach in complex domains likeATM is innovative and challengingsince the exploitation of designpractices and methods fromdi sc ip l ines d i f fe rent f romengineering and human factors islargely unexplored. Innovation insuch domain is constrained by aproblem solving view, neglectingother factors like aesthetic, affective,cultural and emotional aspects ofhuman cognition. CREA! leveragescreativity in action in artisticdomains and offers a process modelto integrate a variety of suggestionsinto integrated solutions.

Enhancement of AGT Telecommunication Security using Quantum Cryptography


ENHANCEMENT OF AGT TELECOMMUNICATION SECURITY

USING QUANTUM CRYPTOGRAPHY

Ecole nationale supérieure des telecommunicationsGET & LTCI-UMR 5141, Paris, France.

Supervised by Marc Brochard.

1. Introduction

The objective of the study is related to the security of Air-GroundTelecommunications (AGT) in the dangerous after 9/11 world where onemay expect serious threats to aircraft safety. We may be concerned by attackon confidentiality, integrity and availability of telecommunications. A wrongmessage or the absence of message may have strong consequences foraircraft safety. Eavesdropping attempts may inform ill-intentioned actors.These hard facts plead for a permanent search of a maximal AGT security.

2. Security for ATN communication

Communications are handled by the Aeronautical TelecommunicationNetwork (ATN) that we discovered during this project. The ATN isincrementally built using existing networks. The security of ATN is a crucialmatter. Aircraft Communications And Reporting System (ACARS) Data Linkmust be secured. Inter Domain Routing Protocol (IRDP) must be securedtoo. Airlines companies require secrecy too. ATN may be secured usingclassical cryptography providing so-called cryptographic security. That meansthat the security relies on the assumed but unproven intractability of somemathematical problems related to prime numbers or elliptic curves.

Quantum Cryptography (QC) provides unconditional security relying on thequantum physics law. Such a security is called information theoretic securitybecause it is proved using the theory of information of Shannon.

The ATN is an Internet network and may switch to IPv6 in the future inorder to provide IP addresses to all equipments. Security and confidentiality inthe ATN will be handled using classical public key cryptography. But publickey cryptography is not proven to be unconditionally secure. No one canclaim that heuristics do not exist to break Public Key Cryptography with highprobability. The birth of Quantum Computers would be the death of publickey cryptography. If Quantum Computers are built in a few years, then publickey cryptography would be dead. Quantum Computers support efficient

CARE INO


algorithms, Shor’s algorithm for instance, to solve the mathematical problemson which public key cryptography relies.

Public key cryptography necessitates a Public Key Infrastructure (PKI). PKI areheavy hierarchical administrative tools. Any security failure in one elementcompromises the security of the system. Thus, PKI is likely to be managed inwell-trusted operator’s areas. PKI will increase the overhead on the band-limited channels. For example, a classical X.509 certificate is about 20Kb.Another typical element of PKI is the Certificates Revocation Lists (CRL),which are very large and must be dispatched to all parties.

Any solution for improving security must be done inside the framework ofthe ATN. It must consider costs and existing infrastructure into account.Existing infrastructure must be re-used. Moreover, any proposed solution forusing QC to secure the ATN must be incremental.

3. Quantum Cryptography - QC

QC is an emerging technology that could, in a few years, provide a totallysecure Internet architecture. ENST is currently involved in the European ISTproject SECOQC which aim is to design specialized Internet optic fiberarchitecture and protocols based on QC. Quantum Cryptography proposesan alternative and a complement to classical Public Key Cryptography. Thepoint is Quantum Key Distribution (QKD) that allows a totally securetransmission of an encryption key. Another possible application field is (air)free space telecommunications which uses faint pulses laser beams. Workshave been conducted in Europe and USA with significant results.

Example of ATN Session. When an Airborne End System (AES)wants to communicate with an A/G (Air/Ground) Application atGround Station (GS), e.g. the Controller-Pilot Data LinkCommunication (CPDLC) Application, AES and GS willcooperate to execute a basic scenario:Step 0: Initialization of ATN’s PKI services for ATN entities whotake part in secure communications such as AES, ContextManagement Application (CMA), CPDLC Application.Step 1: AES creates a CM Logon CPDLC Request and sends itto CMA.Step 2: CMA sends a CM Logon CPDLC Response back to AES.Step 3: AES and CPDLC Application compute a common secretSession Key.Step 4: AES and CPDLC Application protect messages by usingthis Session Key.



QKD allows two endpoints to share a secret key. This encryption key is usedwith an unbreakable encryption algorithm, such as Vernam (one-time pad)cipher, to encode the communication.

The main QKD protocol named BB84 is fully described in the report and avisual implementation is given as a demonstrator. QKD uses a classical openchannel and a quantum channel which may be an optic fiber or a free spacefaint pulses laser beam or any physical device able to transmit unalteredquantum states. The security is guaranteed by Quantum Physics laws insteadof unproven mathematical assumptions: the Heisenberg’s uncertainty principleand the Non-cloning theorem. With QKD, any eavesdropper (spy) can bedetected because its measures perturb the quantum states.

QKD relies on quantum equipments and specialized algorithms. Quantumtechnology is quickly evolving, mainly thanks to the SECOQC project. Whenwe wrote the proposal, the maximum distance obtained with optic fibertechnology was 70 km. Nine months later, it is 130 km and many experienceshave been done to secure Internet links. With QKD, we have two publicchannels: a classical channel to transmit ordinary bits and a quantum channelto transmit quantum states. Both channels are public and used to distill acommon secret encryption key that is used to establish a securedcommunication.

Security is based on secret sharing, either a secret algorithm or asecret key to be used with public encryption algorithms.

A secret can be shared by physical means. E.g.: to use thearmy to share a key between White House and Kremlin.

A secret may be shared by algorithmic means. That is PublicKey Cryptography.

A secret may be shared by quantum means. That isQuantum Key Distribution.

CARE INO


Free space QKD uses faint pulses laserbeam. Table on the left shows theprogress. Due to turbulences in the first1 km atmosphere, 2km Ground/GroundQC is equivalent to 300 km Ground/SpaceQC. Theoretical results with the 2003experiments allow a 1600 km distance forGround/Space QC. Thus, we can imagineQC based on a satellite network. It is800 km LEO satellites. Embedded payloadis 3 to 5 kg, 10 to 30 cm optics. On Earth, it uses a 50 to 100 cm optics. Thesatellites network depends on the payload: from 7 to 43 satellites.

Year Distance Where

1989 32 cm IBM USA

1996 150 m Baltimore, USA

1998 1 km Los Alamos, USA

2000 1.6 km Baltimore, USA

2001 1.9 km QuinetiQ, UK

2002 10 km Los Alamos, USA

2003 23.4 km Munich, Germany

Free Space QKD distances.

Thus, QC can achieve unconditionally secured communications links overrestricted distances depending on the used technology. Big progresses aremade and other alternatives to BB84 are studied: quantum continuousvariables (QCV) and entangled photons (EPR pairs).

We may assume that the distances will be enlarged. In the report, weassumed that all foreseeable QKD technologies have been developed. Forinstance, we assume Free Space Satellite QKD that had not beenexperimented. We looked at the incremental insertion of QKD in the ATN.That is to say that we looked where PKI can be locally replaced by aQuantum Confidentiality Key Infrastructure (QCKI) which would beresponsible of providing confidential sharing of encryption keys between twoendpoints.



4. Quantum Key Distribution

Very efficient encryption algorithms exist and some are proved to beunbreakable by Shannon’s theory of information. For instance, Vernam cipher,also called the one-time pad, assumes that the two endpoints share a key aslong as the message to be encrypted. Vernam encryption is just doing anXOR, i.e. addition modulo 2, between the key and the clear message anddecryption is just doing an XOR between the key and the encryptedmessage. Reading the encrypted message does not give any informationabout the clear message. However the required length of the key, and thefact that the key must be changed after each use, rule out Vernam cipher foran everyday usage.

Modern Data Encryption Algorithms (DEA) such as DES, 3-DES, AES, andelliptic curves cryptosystems allow secure encryption using a fixed-length key.They are considered as unbreakable. But all these algorithms assume that akey is shared between the two endpoints. Thus, security is a problem of keydistribution.

4.a. Classical Key Distribution

Nowadays, key distribution can be done using Public Key Infrastructure (PKI).This will be the case for the Aeronautical Telecommunication Network. A PKIis a security system for the management of keys using asymmetric encryptionalgorithms. But asymmetric encryption is subject to serious attacks with bruteforce, with progress in mathematics or with the possible creation of quantumcomputers. An encrypted message now currently unbreakable may be brokenin ten years, or tomorrow, delivering a posteriori secrets.

Moreover, in the general case, PKI assumes many trustable third parties. Allthis is good enough for most of applications where there is no big business orindustrial stake, no far future concerns, and when national security is notinvolved. For instance, one may admit the PKI system when it distributescertificates and keys for software download or for restricted electronicpayment. But recent affairs12, involving the Echelon electronic communicationssurveillance systems have proved that governments do not hesitate usingmilitary power to serve their own private companies. In recent UNO disputeon Irak, one has learned that the same techniques have been used bygovernments against UNO and opposite diplomacy. Perfect classical digitalconfidentiality needs huge organization and means. Quantum Cryptographymay provide a solution.

The questions are: do we trust encryption algorithms, which are potentiallybreakable? Which PKI can we trust? Even trustable, your PKI may not be

12 http://news.bbc.co.uk/1/hi/world/europe/820758.stm

CARE INO


secure enough since a single break in such a complex system opens a largebreach in the security.

4.b. Quantum Key Distribution

Quantum Key Distribution allows two endpoints to share a key with totalconfidentiality and to use symmetric encryption algorithms. S. Wiesnerdescribed the idea in the 70s. He officially published in 1983. It has been fullydeveloped and finalized by Gilles Brassard and Charles H. Bennett in 1984and it is known as the BB84 protocol [2].

BB84 Basics

The quantum law underlying QKD is Heisenberg principle of uncertainty: twonon-commuting observables of a quantum system cannot be both accuratelymeasured. It ensures that it is not possible to clone a quantum system (no-cloning theorem). Otherwise, it would be possible to measure oneobservable on the original and the other observable on the clone.

The BB84 protocol is simple enough to be understood by a non-specialist ofquantum physics. Photons can have a rectangular or a diagonal polarization,two non-commutable observables. A physical device can observe rectangularor diagonal polarization but not both. Rectangular polarization can behorizontal noted “” or vertical noted “”. Diagonal polarization can be leftnoted “” or right noted “”. Moreover, if a physical device tries to measurediagonal polarization on a photon that is rectangularly polarized, then it gets arandom result: either left or right, each with a probability of 50%. And the actof measurement changes the state of the photon to be the result of themeasure. The situation is symmetric if a physical device measures rectangularpolarization of a photon that is circularly polarized.

Session keys are made of bits, 0 or 1. We agree that: bit 0 can be encodedeither by a horizontal () or a left () polarization of a photon and bit 1 canbe encoded either by a vertical () or a right () polarization of a photon.Such an encoded bit is called a quantum bit or qubit. Transmitting a keybecomes transmitting a sequence of polarized photons.

BB84 Key Exchange

Alice and Bob are connected using two channels. The first is the quantumchannel, typically an optic fiber. The second is the classical channel, typically anInternet link.

1. First, Alice generates a random sequence of bits called the raw key.Randomness is crucial. For each bit, Alice randomly chooses to encodeits value using either the rectangular basis or the diagonal basis. And shesends the photons, one after the other, to Bob using the quantumchannel.



2. For each received photon, Bob randomly chooses to measure it usingeither the rectangular basis or the diagonal basis. Because Alice and Bobchoices of bases are random, the probability that they use the same basisfor a given photon is 50%. If they use the same basis for a given photon,then Bob gets the right encoded bit with a high probability. If they do notuse the same basis, then Bob gets a random result.

3. Then Bob uses the classical channel to tell Alice which bases he used forthe measurements. And Alice, also using the classical channel, answerswhich bases are correct according to her own encoding choices, i.e.when they used the same basis. These communications are public.

4. When they used the same basis, the bit encoded by Alice is identical tobit decoded by Bob with a very high probability. They get a sharedsequence of bits, called a sifted key, which that be used to build a sessionkey. The length of the sifted key is about half the length of the raw key.

Example. In Table 4-1 below, rectangular and circular bases are representedby symbols “⊕” and “⊗”. The first line ARK contains Alice’s randomly chosensequence of bits. Second line ARB contains the encoding bases randomlychosen by Alice for each bit and the third line AQB contains the qubits, i.e.the photons with the appropriate polarization. Fourth line BRB contains Bob’srandomly chosen measurement bases and fifth line BQB contains the resultsof the measurements. We have put a symbol “?” to mention that Bob’smeasurement has a random result which will be discarded anyway.

ARK 0 0 1 0 1 1 1 0 0 1 0 0 1 1 1 1 0 1ARB ⊕ ⊕ ⊗ ⊕ ⊗ ⊕ ⊗ ⊗ ⊗ ⊕ ⊗ ⊗ ⊕ ⊗ ⊕ ⊕ ⊕ ⊕AQB BRB ⊕ ⊗ ⊕ ⊕ ⊕ ⊗ ⊗ ⊗ ⊗ ⊕ ⊗ ⊗ ⊕ ⊕ ⊕ ⊗ ⊗ ⊕BQB ? ? ? ? ? ? ? BSK 0 0 1 0 0 1 0 0 1 1 1

Table 4-1 – A QKD Session.

The last line BSK contains the bits for which Alice and Bob have chosen thesame basis, this is the sifted key which value is “00100100111” in ourexample.

Eavesdroppers and Security

The eavesdropper, usually named Eve, has access to both channels. If Eveaccesses a photon, she has no way to know the basis used by Alice toencode the bit. Thus, she has to guess a basis for measuring the photon. Andthen she resends the photon to Bob. This is the intercept-resend strategy. Ifshe chooses the same basis as Alice for measurement, then she gets the right

CARE INO


value and the resent photon is in an appropriate quantum state. If shechooses the wrong basis, then she destroys the quantum state of the photonand, in the cases where Bob chooses the right basis, he gets an incorrectresult in 50% of the cases. On the average, Eve chooses the wrong basis in50% of the cases. Thus, Eve’s action introduces a supplementary error rate,about 25%. In this case, Alice and Bob can detect the intrusion and know thatthe sifted key cannot be trusted.

Another strategy for Eve is the man-in-the-middle attack. In this attack, Evegets control over the two channel and lets Alice think she is communicatingwith Bob and conversely. Eve plays the role of Bob w.r.t. Alice and plays therole of Alice w.r.t. Bob. In this case, one must rely on classical authenticationalgorithms stemmed from classical cryptography or recent quantumauthentication algorithms.

Many articles give a rather complete description of the non-impossiblequantum attack strategies, for instance beam splitting scheme orentanglement scheme or quantum copying scheme or collective attacks, invarious configurations and for various QKD technologies, and why they areunlikely to succeed. Formal proofs of security rely on protocols such as thefollowing BB84. They uses Shannon’s Information Theory and, mostimportant, the laws of Quantum Physics.

BB84 Protocol

The BB84 protocol is used over the physical devices handling the keydistribution. Rationales for this protocol are multiple. First, the quantumdevices, for producing quantum states, for transporting and measuring them,are not totally perfect. For instance, one must consider the dark count that isthe probability of detection of an unsent photon, a low probability about 10-5

that cannot be neglected according to communication standards. One mustconsider the probability of measure errors due to apparatus defects that is farmore important. Thus, if the sifted key length becomes a few percents of theinitial bits string length, it can be considered as a performance. The protocol isthere to take into account the error rates due to technical imperfections andto the eavesdropper’s action. The error rate in the sifted key is called theQBER for Quantum Bit Error Rate. The aim of the protocol is to reduce theQBER to standard communication Bit Error Rate (BER), about 10-9, and toreduce as much as wanted Eve’s knowledge about the key. The steps of theprotocol are the following:

1. Sifting. Alice sends a random string of bits, the raw key, as describedabove. Alice and Bob must be synchronized to detect photons that Alicedid not send but Bob received and, conversely, photons that Alice sentbut Bob did not receive. The result is the sifted key. The length of the



sifted key is about a few percents of the length of the raw key. At thisstep, Alice and Bob may detect Eve’s intrusion because a significantintrusion must raise the usual error rate.

2. Reconciliation. The sifted key is made of qubits on which Alice and Bobagree because they have used the same encoding basis. However, somebits may differ because the quantum apparatus is not reliable or becausethere has been a light intrusion of Eve which has not been recognized asso. The error elimination algorithm uses the public classical channel.Several algorithms have been proposed. For instance, it has beenproposed that Alice and Bob use the same random permutation of bitsto randomize the locations of errors. Then, the key is divided into smallenough equal-size blocks such that one block is unlikely to contain morethan one error. Alice and Bob compare the parities of their respectiveblocks and discard blocks for which parities differ. After reconciliation, thesifted key may have been shortened but it is almost certainly sharedbetween them.

3. Privacy amplification. It may be that Eve knows some bits of the keyresulting from the previous operations. Privacy amplification is atechnique to reduce Eve’s information. The price is once again shorteningof the key. Again, several algorithms are possible. For instance, Alicerandomly chooses two bits and tells Bob the position of these bits. Aliceand Bob replaces the two bits by the result of their XOR. If Eve has onlypartial information on these two bits, i.e. if she knows zero or one bit,then she has no information on the XOR result. Therefore, Eve’sinformation is less than before. Alice and Bob to reduce Eve’s knowledgemay repeat this operation.

4. Authentication. The two parties identify themselves. This may rely onclassical algorithms not especially related to Quantum Cryptography orto recent Quantum Authentication algorithms developed by the authors.These algorithms assume that a piece of data, an authentication key, isshared by Alice and Bob before all. In fact, they may share a stack ofauthentication keys. They are subject to keys exhaustion, and then Denialof Service (DoS), if an eavesdropper simulates a lot of connections. Weproposed a new algorithm, which protects itself against keys exhaustion.At the difference of usual approach, his algorithm is dedicated toQuantum Cryptography.

Then Alice and Bob share a key with a very high probability and Eve’sinformation about the key is as small as wished.

CARE INO


A single QKD Link

The simpler network consists of a QKD link between two enclaves; seeFigure 4-1, which marries QKD with classical Internet security protocol IPSec.QKD is used for key sharing between two enclave gateways.

Figure 4-1 – A single QKD Link

The enclave, a Local Area Network (LAN), is assumed to be secured. An IPSecsecured Internet link connects the two gateways. IPSec is a well-establishedInternet technology that allows traffic between two endpoints to beconfidential provided the endpoints share an encryption key. The twogateways ensure the routing of IP communication. The only non-classicalfeature is that the keys necessary to IPSec are distributed using quantumtechnology. The two QKD devices produce continuous streams of bits, whichcan be used for regular key renewing.

A long QKD Link

Simple QKD links as above are limited to several tens of kilometers length. Inorder to extend the length, one may use QKD data relay. One must notethat a QKD data relay is not a quantum repeater. A QKD data relay is anetwork apparatus able to establish a single QKD link with the previouselement of the chain and another QKD link with the following element of thechain. It is a data relay with the following characteristics:

Relay k establishes an encrypted communication (a QKD link)with relay k-1.

Relay k receives encrypted data from relay k-1.

Data are decrypted and stored in the memory of relay k.

Relay k establishes an encrypted communication (a QKD link)with relay k+1.

Data in memory are encoded and sent to relay k+1.



We can see that QKD data relays present a serious weakness: data appearunencrypted inside the relay memory. QKD data relays establish pair-wisesecure communications using QKD in order to securely transport a randomlygenerated encryption key, hop-by-hop from one endpoint to the other as inFigure 2. The QKD relays network at the bottom of the figure is used toexchange an encryption key that used to encrypt the communication on thetop Internet link.

Figure 4-2 – QKD relays used to transport keys.

The communication between QKD relays is done as the communicationbetween Local Area Networks (LAN) enclaves of section 3.a above. Theencryption key that is exchanged using the QKD Relays Network appearsunencrypted inside the relays. Thus, the relays must be seriously protectedagainst eavesdropper. In Europe, due to the concentration of cities, such ascheme could be used by many institutions. This may not be applicable tolarger countries such as USA, Canada or Russia where extended non-urbanareas exist.

Quantum Networks

In previous section, it was chosen to use QKD Relays Network to transmitkeys between two endpoints of an ordinary Internet network. But the QKDrelays may be used to transmit the plain traffic. Then, we obtain a trueQuantum Network. In this network, we have a set of enclaves. Some of theseenclaves are pair-wise connected by a Single QKD Link. It is not obvious thatsuch a network would be more interesting. Security level is the same. And itwould be a little more complex to build and to administrate without the gainof a wider usage.

CARE INO


5. Applying Quantum Cryptography to ATN

QCKI can be introduced locally to secure a sub-network of the ATN withoutaltering the whole structure of the ATN or the PKI system. The sub-networkis unconditionally secured and communicates with the outer world withclassical gateways. For instance, one can think of securing a big airport withoptic fiber technology or securing an A380 aircraft with the same technology.We get QCKI Islands inside the ATN.

QCKI Links and QCKI Islands.

QCKI can be introduced locally to secure links between ground entities of theATN provided that constraints of distance are respected, now 130km. Forinstance, we could secure links between all airports of Aéroports de Paris(ADP) or between ground stations. If the distance is more than required, onemay think of using satellite QC although the technology is not ready.Otherwise, the aim of the SECOQC project is to build optic fiberunconditionally QC-secured terrestrial dedicated networks. This technologymay be used to secure the ground part of the ATN and to replace PKI.

A concept developed by ENST is that of QBONE. One may think of aclassically secured network such as the ATN or a bank network. Let usassume that this network must have Access Points located outside of itssecurity zone. For instance, an ATM machine must be connected to the banknetwork but it may be located in an unprotected commercial center. For theATN, the external AP could be the aircrafts. Communication with the aircraftcan easily be monitored, thus we cannot assume secrecy.



The QBONE.

Let us consider aircraft as external AP to the classically secured ATN. AGTData Link (DLK) provides numerical communications between groundstations and aircraft. They are used for Graphical Position Reports, ContactReports, etc. One may classify different threats:

Monitoring. A third party may listen to the DLK communications and gaininformation on the traffic. Current DLK communications do not guaranteeprivacy.

Spoofing. A third party may listen to the DLK communications and gainauthentication information in order to impersonate one of the parties.

Modifying. A third party may impersonate the second party with respectto the first party meanwhile he may also impersonate the first party withrespect to the second party (man-inthe-middle attack). Integrity of thedata is not preserved. Data may be corrupted.

It is very easy to monitor Aircraft Communications Addressing And ReportingSystem (ACARS)Data Link Messages. One needs a personal computer, asound card, a Radio Frequency (RF) scanner and few software freely availableon the WEB.

Thus, the need to secure aircraft communication with ground stationsappears clearly. We consider them as AP to the ATN. Free space QCKI canbe used to distribute encryption keys:

To aircraft entering the European sky either from the ground if controllersoblige the aircraft to cruise at the vertical of one of some chosen points atthe frontier of Europe; or from satellites otherwise.

CARE INO


To aircraft standing at the airport, maybe not wired to the airportterminal. The control tower could securely distribute a key to any aircraftstanding on the tarmac.

Then the encryption keys are distributed to the ground stations using theclassically secured ATN.

6. Perspectives

The perspectives of the work performed could be a coupling of AirIdentification Tag (AIT) developed by the university of Graz and Eurocontrolwith QKD. AIT is the watermarking insertion of flight identification in VHFpilot-controller communication. Any party duly equipped can see the otherparty identification on a special visual device or, in the case of controller; itcan be used to highlight the speaking aircraft on the radar screen. AIT did notintend to guarantee authentication of the parties. AIT has been designed toreduce the controller workload and stress. Authentication and integrity canbe obtained by cryptographic signature technology provided that the twoparties share a key. Free space QKD is used from the control tower todistribute a key to aircraft standing at the airport. The AIT message couldinclude the flight identification, the current GMT Time and a signature of bothprovided by one of the hash functions of the classical cryptography cookbook.

The report of our work also describes more ambitious scenarios based ondifferent QKD techniques to secure the whole ATN while respecting thecriteria of incrementally of the insertion of QKD inside a PKI-based systemand the criteria of complementarities of the two techniques. The mostambitious plan would be to use satellites-based key distribution. The requirednumber of satellites varies from 7 to 43 depending on technology evolution. Itis a costly solution that may be used only if PKI is broken one day byQuantum Computers or mathematical progress.

All project information and report are available from the CARE INO web siteat the following URL :

http://www.eurocontrol.int/care-innov/public/standard_page/innov2_quantum.html

A full description of the possible applications of QKD to ATN may be foundin the report mentioned above.

VisuAirport : Adapted Observation for Activities of an Airport


VISUAIRPORT

ADAPTED OBSERVATION FOR ACTIVITIES OF AN AIRPORT

ARMINES, Ecole des Mines, Paris, France. Readymade, Paris, France.

Supervised by Marc Bourgois,.

1. Introduction

As stressed by the European Commission in its “vision for 2020” for Aircraftand an air transport system, and by the Advisory Council for AeronauticsResearch in Europe (ACARE) in its Strategic Research Agenda, volume1(SRA1), inevitable increase of traffic shall not be accompanied by an increaseof accidents.

The digital revolution is enabling huge strides to be made in aircraft design,production, manufacturing, maintenance and operating and trafficmanagement.

The technical content of SRA1 is driven by five major challenges that interactin addressing the top-level objectives. The ambition to provide moreaffordable, cleaner, safer and more secure air travel determines the majorchallenge areas.

The challenge is concerned with the safety of the future air transport systemcharacterized, as it will be by five important dimensions:

A three-fold increase of the “density” of the system through increasedtraffic;

All weather operation;

99% of flights departing within 15min of schedule;

Operations at Airports 24 hours per day;

New systems of flight management.

Synthetic 3D, enhanced man machine interface, and real time assistance canprovide some contributions and solutions in the efficiency and safety fields. Inthis context, a new way to observe the activities of airport is proposed:adapted observation for activities of an airport.

CARE INO


2. Objectives and innovative ideas of the study

The adapted observation for activities of an airport aims at providing severalairport actors with a realistic 3D vision of their working environment, tunedto their needs like the sight of the airport, the landing strips, the circulationand parking areas, the airport infrastructure, the aircrafts and any othervehicles operating on the airport. The main interested actors are the groundoperators, and more especially those in charge of the aircraft and airportoperations.

The idea is thus to propose for any actor a 3D vision giving instantaneous andreal time synthetic view of any area. The airport environment not being fixed,all the vehicles and aircrafts must be localised in real-time, thanks tolocalization sensors (like GPS or others). Thus, it will be possible to providesynthesized images very similar to the real sight and updated in real-time. Thedisturbing influences, such as bad weather conditions, buildings and any otherelement obstructing or reducing visibility would be eliminated.

The principal innovation is to exploit a new type of adaptable and multi-functional visual interface, using desktop computers as fixed stations or tablet-PC as mobile working stations. By associating this mobile station with a real-time modelling framework of the environment, relatively well structured andknown, we are proposing to the airport actors a new observation concept oftheir working environment. The use of the new inertial sensors, borrowedfrom the virtual reality techniques, able to provide three degrees of freedomin real-time, could be used to easily choose the desired point of view, bymanipulating the tablet-PC, localised by an inertial sensor: the observer, whileturning relatively the tablet, changes the view point’s orientation in a pseudo-natural way. The user can easily choose, without a keyboard, only by using thepen and the interactive screen, the observed area, the display functionalities,the memorized configurations of several points of view, to communicate withother people, etc.

Tablet PC and inertial sensor used for visual interface in mobile station



3. Demonstrator

In order to demonstrate the added value of applying such technologies in theairport operation field, a demonstrator was developed and used to supportsome experimentation with real users.

The demonstrator developed in the context of this work, provides the userwith a 3D visual interface with the possibility to change his/her point ofobservation and to configure a set of viewpoints. The first step of the projectincludes an analysis of the observation situation in the airport area, thespecifications for the visual interfaces and for the scene modelling software.The second part of the project comprises the airport 3D modelling(geometrical, visual and kinematical behaviour), the aircrafts operating in thearea of interest as well as the vehicles. The ground vehicles will be supposedto transmit their positions in real-time (unsing standard GPS). Thedemonstrator gives the operator the possibility to change his/her point ofobservation according to his/her needs. He/she can also configure a set ofviewpoints usable at any time he/she wants. Consequently, the third part ofthe project consists in designing the software functions allowing the viewpointmodification, by using a simple mouse, the tablet-PC’s pen or using an inertialsensor connected to tablet PC. The various ways of choosing the viewpointswill have to be tested, improved and validated. The ergonomic aspect of theviewpoint command must be carefully taken into account at the cognitivelevel. It is never easy to handle by rotation a direction (here the view point’sdirection), because rotations (according to the three orthogonal axes) are notcommutative operations. The use of an inertial sensor will be the solution fora pseudo-natural behaviour of the observer.

For the final functioning system, one needs a synchronous communicationbetween all the visualization stations (fixed or mobile), transmitting all thevariable parameters of the observed scene (flight parameters of the aircrafts,vehicles and people positions on the ground). For the demonstratorproduction, carried out during this one-year project, no development on datasynchronization is required.

The software development has been carried out starting from the softwareplatform of the French company Virtools, which was designed for applicationsof animation and virtual reality. The application will be tested with a visualinterface on a desktop station (micro-computer PC) and with a visualinterface on mobile station (tablet-PC) with wireless communication betweenthem.

CARE INO


4. Creation of synthetic views

Needing to provide the most realistic environment display, the syntheticviews have been based on the model of Roissy airport area: thanks toAeroport De Paris (ADP) and Roissy Charles de Gaulle officials for their kindauthorisation granted to the project. The demonstrator is providing specificareas of the airport (parking and boarding gates) and all the vehicles andaircrafts operating in these areas.

The development of the 3D display has been made with mass plan of theinfrastructure, pictures taken on the airport platform and ground vehiclemanufacturers. Several tools have been used such as 3DStudio Max, Maya,Virtools.

In order to get a better realism, vehicles textures have be added based onpictures taken on the airport platform.



5. Evaluation

Considering the short project duration (one year project), only 2 scenarioshave been developed for the evaluation.

The first scenario was focussing on aircraft/ground vehicles accidentprevention on the parking area focusing on communication and coordinationimprovement between ground operators and aircrew.

The demonstrator is based on a Roissy airport parking area where visibilitybetween ground vehicle and aircraft taxiing might be obstructed or reduced.The airport parking area synthetic view is provided to the ground vehicle’sdriver on a tablet-PC. The synthetic view provided corresponds to the realview of the driver.

Two types of incident/accidents are considered for demonstration: crossing ataxiway with no obstacle, and crossing taxiway with building obstructing thevision.

In the first case, aircrafts are running in permanence on the taxiway when aground vehicle is supposed to cross this taxiway. There is nothing which couldobstruct the vision, but this vision is drastically reduced in case of bad weatherconditions (fog for example). In order to avoid risk of collision especiallywhen visibility is getting bad, a synthetic view is provided to the groundvehicle with an alert system providing collision warning message. This collisionwarning information may be a sound alert and/or a synthetic red light (2D or3D display).

Example of synthetic display of specific parking area at Roissy Charles Gaulleinternational airport.

CARE INO


In the second case, aircrafts are running in permanence on the taxiway whena ground vehicle is supposed to cross this taxiway, but a building isobstructing the vision of the ground vehicle and the ground driver might notsee aircrafts running on the taxiway and/or other ground vehicles. Wheneverthe building is eclipsing an aircraft running on the taxiway, or an other groundvehicle, it becomes transparent on the tablet-PC. This enables to have betterinformation awareness and shall avoid ground collision in case of bad visibilityconditions.

The second scenario was focussing on communication and coordinationimprovement between airport operators. A user (the “coordo”) is operatingaircraft stopover on the airport parking area while another user (thesupervisor) is in charge of coordinating all the aircrafts during their stopoverson the behalf of an airline in the supervision room. The synthetic parking areavisualisation is provided to both users, on tablet-PC for the “coordo” and ondesktop PC for the supervisor. The synthetic view is providing complex anddetailed vision of ground staff and vehicles operation. This view is the samewhatever weather conditions are, and at any time of the day (be it day ornight). On top of the synthetic view, computer assisted tools are providedlike, for example, a small 2D window indicating the stopover progresschronology (refuelling, maintenance, catering, luggage loading or unloading,mechanical interventions, technical checking, etc.), or detailed informationabout vehicle displayed on mouse selection. This increases the mutualunderstanding of the stopover processing.

All project information and report are available from the CARE INO web siteat the following URL:

http://www.eurocontrol.int/care-innov/public/standard_page/innov2_visuairport.html

SCOPE : Safety of Controller-Pilot Dialogue


SCOPE

SAFETY OF CONTROLLER-PILOT DIALOGUE

Thales Research & Technology France.

Thales Corporate R&D centre, Orsay, France.

IntuiLab, man machine interfaces SME,Toulouse, France.

IRIT, Institut de Recherche en Informatique, Toulouse, France.


1. Introduction

The reliability of communications between pilots and air traffic controllers isof paramount importance to air traffic safety. As such, the detection ofcommunication errors between pilots and controllers has always been amajor safety issue. Many of those errors arise from undetectedmisunderstandings between the pilot and the controller during radioconversations. An automated tracking of the pilot-controller dialogue, in orderto check the matching between clearances (controller) and acknowledgments(pilot), coupled with a verification of the effectiveness of the modification offlight parameters would dramatically improve flight safety.

The purpose of the SCOPE project was to increase the reliability of pilot-controller communications using automatic speech recognition, in order totrack the dialogue between the controller and the pilot, and multimodalinformation presentation, in order to present the results of the tracking to thecontroller through an enhanced interface. Air Traffic Control (ATC) systemsrequire a high level of reliability and are subject to real-time constraints. Withsuch a challenge in mind, the SCOPE study consisted in selecting andmodelling a relevant subset of ATC phraseology and exploring the potentialof the use of voice recognition for pilots and controllers as well as theconditions of its successful implementation.

2. Automatic speech recognition selection

The first task of SCOPE consisted in identifying and analysing the mostappropriate tool for automatic speech recognition (ASR) in ATC, with a

CARE INO


special focus on availability, robustness and ability to be incorporated intoATC systems.

The characteristics of ATC communications (controller-pilot dialogue) are thefollowing:

Constrained domain-specific language (recurrent utterances, use ofcallsigns) and limited vocabulary;

Important variability of speakers (potentially bad accents) with often noavailability for training;

Noisy environment (cockpit, control rooms) and poor quality transmissionchannels (radio communications);

Stressed speech: rapid delivery, bad pronunciation, interrupted oroverlapping utterances.

In more technical terms, automatic speech recognition of ATCcommunications will face the following constraints:

Limited bandwidth (300 to 3300 Hz) of VHF voice com; English languagehas major energies in higher parts, i.e. “th” (8 to 9 kHz);

Technical instability at beginning of speech (VHF);

Incomplete words (phonemes) a the beginning, i.e. “fthansa one two”(Push To Talk switch);

Spontaneous speech with repetitions, hesitations: “mmh”, “haa”(approximately 4 to5% of the utterances);

Mainly non native English speakers;

Mix of official ICAO languages (English, French, Spanish, Russian), i.e.“Geneva” in approximately 16% of the French utterances;

Pronunciation of navigation point names in speaker’s mother tonguelanguage;

Special keywords: break, disregard, correction;

Mix of English with native language courtesy forms, i.e. “guten tag”,“gruetzi”, “bonjour”.

Therefore, the most important criteria for the selection of an ASR system tobe used to track such communications are the following:

Possibility to build domain-specific language models or grammars and usethem dynamically at runtime, in order to select the specific sub-grammarwhich is appropriate for the recognition of a sub-dialogue; moreover, if noreadily usable corpus of ATC communications is available, formalgrammars will be preferred to statistical language models;



Speaker independence to cope with speaker variability, as well as limitedand rapid training if not absence of training at all;

Noise robustness through noise processing capabilities;

Output of type n-best or lattice for easier error repairing and dialoguemodeling capabilities in order to process complex dialogue structures.

Given a selection of candidates matching these criteria, the final choice of aspecific tool can then be made on the basis of the performances (error rateand processing time) of the different systems through experimentations andtests. These tests can be performed directly on ATC communication samplesor else on corpuses from similar domains (military communications forexample).

Given the nature of the targeted application, that is ATC communicationswith limited vocabulary and constrained language, and the associated strongconstraints (noisy environment, low quality transmission channels, overlappingutterances and stressed speech), Nuance 8.0 from Nuance Communicationsappeared to be the appropriate choice. Nuance is perfectly designed forlimited vocabulary, and its grammar construction, edition and integrationfacilities make it readily usable for a constrained language such as the oneused by controllers and pilots. The results obtained with Nuance 8.0 on sucha language are much better than those obtained with other ASR software,and Nuance is able to perform recognition for this kind of application in real-time.

3. Air-ground communication grammar modelling

The next task concerned the modelling of the pilot-controller communicationlanguage in order to build the grammar for the selected recogniser. One ofthe most difficult aspects of implementing ASR is the creation of the grammarfile. The terminology of all possible phrases must be rigidly defined. As such,callsign recognition is the main step to achieve in the ATC domain. Withoutthis information, the recognition process cannot fulfil the requirements ofATC systems. Therefore, a formal model of English callsigns was proposed. Allpossible pronunciations were formally described, in order to limit the size ofthe generated grammar and optimise speech recognition. The methodologyused in the SCOPE project for grammar constraining made it possible to useonly a small grammar for callsign recognition. In particular, the grammar wasadapted to a limited list of flight plans for the controller’s current sector. Thissolution ensured that Nuance 8.0 together with the SCOPE grammar forcallsign recognition could perform clearance/acknowledgement recognition inreal time and yielded excellent results under some conditions (average Englishaccent, good experience of the OACI alphabet and of company names,average voice volume and speed).

CARE INO


4. Demonstrator

The third task focused on the design of a multimodal interface for thepresentation of the speech tracking results to the controller, and thespecification of the underlying architecture. Several scenarios were proposedin order to demonstrate the usefulness of ASR in ATC communications. Theretained scenario coupled:

Verification of aircraft identity using matching of the callsign obtained fromASR of the controller’s clearance and the callsign obtained fromwatermarking on the pilot’s acknowledgement;

Alerting of clearance/acknowledgement conflicts using comparison ofrecognition results for both the controller’s clearance and the pilot’sacknowledgement.

As a consequence, the multimodal interface was designed as to present alertsto the controller in case of mismatches or conflicts. The resulting softwarearchitecture includes two ASR tools, one for the controller and one for thepilot, plus a radar image supporting presentation of multimodal information(twinkle and plug-in) together with an air traffic simulator in order to betterillustrate the scenario, see figure below.

Air traffic simulator

Twinkle +

Plug-in

Controller A.S.R.

Nuance vocal recognition

engine

Air traffic control

specialized Grammar

Ivy Bus

Pilot A.S.R.

Nuance vocal recognition

engine

Air traffic control

specialized Grammar

The last task consisted in implementing a demonstrator of the tracking systemwith a multimodal interface according to the specifications. The resultingdemonstrator consists of two workstations with ASR capabilities; one for thecontroller’s working position, which includes also a laptop running the



multimodal radar interface, and one for the pilot. The demonstrator permitsto illustrate several cases of controller-pilot communications leading toconflicts, which are detected by the SCOPE ASR system and presented tothe controller on the radar image, and thus demonstrate the usefulness ofASR for the tracking of ATC communications. The SCOPE demonstrator willbe presented live during the 3rd Eurocontrol Innovative Workshop &Exhibition.

5. Perspectives

Considering the experimental conditions where voice recognition was usedand demonstrated its potential for ATC, the perspective might be to studythe scaling of the SCOPE approach in order to tackle issues that were out ofthe scope of the initial project, such as stressed speech or limited bandwidth(and operational conditions). The study would rely on a corpus of liveconversations between controllers and pilots and would be dedicated to fineacoustic and language modelling together with contextual and semantic basedrepairing of recognition errors. The efficiency of the resulting ASR anddialogue technology could be illustrated through one of the followingapplications, among others: a reduction of controller’s cognitive load, aredundancy of information for the pilot or a real-time analysis of controller’sworkload.

All project information and report are available from the CARE INO web siteat the following URL

http://www.eurocontrol.int/care-innov/public/standard_page/innov2_scope.html

ANIMS : Improving the Efficiency & Safety of ATM User interfaces with Visual Animation & Sounds


ANIMS

IMPROVING THE EFFICIENCY AND SAFETY OF ATM USER

INTERFACES WITH VISUAL ANIMATION AND SOUND

IntuiLab, Man Machine Interfaces SME,Toulouse, France.

Intactile Design, Graphic design and Sound design SME, Montpellier, France.


1. Introduction

In the past years, the development of tools for air traffic control wasconsidered by all as a task for ATC experts, software engineers, and humanfactors specialists. Meanwhile, the times have changed for the computerindustry. Nowadays, thousands of designers work on interactive software forcars, aircraft cockpits, office systems or games, because they help to makebetter and more usable systems. The role of software engineers fades back tomaking systems that work, not designing their interface. At the same time, asuser interface technology grows more mature, it offers possibilities that canimprove the efficiency, naturalness and even safety of operation of interactivesoftware. This comes at a time when new air traffic management conceptsand tools are flourishing and thus are raising the need of a carefully designedinformation environment for air traffic management operators.

The ANIMS project studies the benefits and conditions of use of two relateddesign-intensive interface technologies: animation and sound. Research andexperience in the last decade has shown how what appeared as futile details,namely interaction styles and visual design, could determine the success orfailure of an ATC system. In the same way, recent research shows that thequality of feedback and alerts, however subtle they are, has a notableinfluence on situation awareness, mutual awareness and safety. Animation andsound are two interaction modalities that share many characteristics: they areintrinsically dynamic modalities (as opposed to graphics, which are mainlystatic), they are time intensive in terms of computer CPU, they introducepotentially complex notions of synchronisation into software architectures,and they solicit specific perceptual and cognitive capabilities of users.

ANIMS is carried out in collaboration between Eurocontrol ATC experts,researchers in user interfaces from IntuiLab, and visual and sound designers

CARE INO


from Intactile Design. The first phase of ANIMS aimed at demonstrating thepotential of well designed sound and animation in HMI for air trafficmanagement. The project also aims at providing the Eurocontrol agency withmeans for popularizing that technology among ATC providers and helping theindustry to do actual work with designers: design guidelines, methods fordesigning and specifying animated and sound notifications or feedback,guidelines for adding them to existing HMI software architecture. During thisfirst year 2004, the project produced both a state of the art anddemonstrators.

2. The state of the art

The state of the art report on animation and sound in Human MachineInterfaces examines their uses in HMI, whatever the application domain. Itreviews the field both from a theoretical point of view (based on scientificliterature review) and from a practical point of view based on current HMIpractices in laboratories and industries. The report first focuses on animations.It proposes a practical definition of what animations are and defines threedifferent types: story-telling animation, system or user triggered animation, anduser driven animation. It then describes how they can be modelled, and whattechniques can be used to create them. The main possible uses are thenlisted before considering human factors, with both advantages and possibledrawbacks of animation. The report then focuses on sound from a moretheoretical point of view, as current experience and research in sound aremore limited in the field of HMI. It defines the sound both as a physicalphenomenon, and as a perceptual phenomenon. It identifies the possible useof sound in HMI following three different possible applications: feedback,alarms and information. It then describes the current use of both sound andanimations in ATC HMI, both in operational systems and in the field ofresearch. The state of the art also contains an extensive bibliography onsound and animations. During this state of the art work, a workshop wasorganized at EEC, with HMI and Human Factors experts from both the fieldof ATM and other domain such as the car and office software industries. Theworkshop both provided insight as to the nature and uses of animation andsound, and fostered ideas about possible cross-domain collaborations on userinterface design.

3. Animation design

After the domain review, operational scenarios for assessing anddemonstrating the potential of animation and sound were selected incoordination with HMI and Human Factor experts from EEC and CENA. Thescenarios were chosen for a medium term implementation, because it



provides a realistic schedule for implementation of new HMI features, andbecause we wanted the intended public (air traffic controllers, ATC systemdesigners) to focus on the use of animation and sound rather than onhypothetical long term scenarios. Five different scenarios were selected. Fourof them are based on the radar view display: enhancing STCA and down-linked ACAS-RA alert, presenting more information on line 0 of a flight label,and notifying which flight is calling (with the help of flight identificationwatermarked in the voice radio channel). The fifth scenario is for a flightsequencer, based on a timeline HMI where either the user or the system canre-sequence flights in the timeline.

Demonstrators were then designed and developed through an iterativeprocess. A methodology has been defined so that involved participants will beable to describe and share their idea on animation and sound design. Due tothe intrinsically dynamic characteristics of both animation and sound,describing them either in draft documents, working documents and of coursein design and specification documents is not an easy task.

For the animations, we extensively used storyboards, in a similar way to themovie, cartoon or game industry. Most of our storyboards were drawn bygraphic designers involved in the design. However, in some cases, technicalpeople involved in the design also drew such storyboards. These were usuallyeven more simplified. This is especially important if the design team is notlocalized at the same location. This was actually the case in the ANIMSproject: Intactile Design and IntuiLab people only met from time to time.Most design meeting were based on phone conferences, with the support ofstoryboards sent by email (after scanning) or by faxes. In this case, it wasimportant that both parties involved in the meeting were able to expresstheir designs. Of course, graphic designers are able to quickly drawstoryboards of better quality.

After some storyboarding iteration, we decided which designs were the bestfor the different scenarios to be demonstrated. The selected animations werethen mocked-up as “Flash Macromedia animations”. This was useful becauseit gives a way to really see the dynamic behaviour of an animation over time.These mock-ups can be viewed with a web browser and a plug-in freelyavailable from Macromedia. They can be put on a web site (in factMacromedia Flash was designed for this purpose) but they cannot beincluded in a documentation file like this one. The following figure just showssome snapshots of such an animation.

CARE INO


Example of a typical storyboard for a movie or a cartoon.

As a conclusion, from the methodological point of view, storyboards and“flash mock-ups” are useful and certainly necessary tools. However, they arenot specification tools. Neither storyboards, nor “flash mock-ups” are preciseand complete enough to serve as specifications. Such specification tools wereout of the scope of this first year ANIMS study which focused ondemonstrating the interest of animation and sound in HMI.

Snapshots of a Flash Macromedia animation mock-up (scenario 4).

4. Sound design

For sound, the problem is even more difficult as it not even possible to drawor write a snapshot of a sound (as it is possible for an animation). Despite



their limitations for that purpose, we have also used storyboarding techniquesfor describing combinations of sound and animation. As we limited our studyto the use of pre-recorded sounds, there was not real need of expertise inthe field of sound computation/generation in the HMI. Thus HMI developerswere less involved in the design process. The main actors involved weredesigners: a graphic and sound designer, a sound designer with a backgroundin music and sound recording, human factor experts, HMI / ATC experts.

Description of a sound.

For the purpose of sharing the sound design among people involved in thedesign process, we needed tools for recording the design. Again, storyboardscan be used during the design process for describing when the sound shouldbe emitted by the HMI. Some features of the sound can also be described.For example shown here which is only a third of the whole “sound”storyboard; the time line at the bottom of the storyboard shows whensounds occur. It also shows that some of the sounds used are repetitive ones.

Because of the complexity of sound inclusion in the course of a scenario andin synchronisation with animations, this storyboard was refined and describedin a large electronic document, has shown in figure below.

CARE INO


Final storyboard used for sound design in scenario 1.

We also needed tools to mock-up some sounds. A very simple and quickway to “sketch” a sound is to sing or whistle the sound. The simplest way to



keep these sketches are to record them with a tape recorder. Sounddesigners also have their own notation system, similar to the score formusicians. We did not use this kind of notation during the design process.

After the step of the sound mock-up, the real sound needs to be created. Forthis purpose, the sound designer used a sound synthesizer. This tool canhardly be used by someone other than a sound designer. Some iterationbased on .wav files can be done, before the final results. The final sound mustthen be inserted in the HMI with precisely and at the right time. For example,in scenario 5a some sounds pre-notify the user that the system is re-sequencing. After some delay, an animation shows exactly the sequenceupdate. The delay between the beginning of the sound and thecorresponding animation must be set with care so that the user has enoughtime for changing his/her focus without having to wait for the animation.

From the methodological point of view, sound should be considered in itsecological context. To envisage the ecological context, we should take intoaccount the sound “landscape” in an operations room. This was not possiblein the ANIMS project due to the limited resources available. One way towork in this direction would be to record these ambient sounds in differentoperational rooms, and to use these recordings as a background for thesound emitted by the HMI. This was done in a very simplified way, by justtaking into account controller and pilot utterances as available in the RADE1data.

5. Prototyping and results

Finally scenarios were prototyped and developed, using the IntuiKitenvironment (an IntuiLab product), in a more realistic context, involving someflight traffic as well as pilots and controller voices. Both traffic and voices werere-used from the RADE1 experiment. The demonstrators both show theseveral benefits of designing animation and sound in ATM user interfaces, andsuggest possible designs for medium term implementation.

Animations are especially useful and should be considered when there is aphysical, temporal, or semantic distance between the cause of an event andthe user focus at the time of the event. There are others situations whereanimations may be interesting, even if it would be more difficult to prove orconvince “classical” HMI designers and developers. Animations can make theHMI more fluent, more natural, with a better prosody. This kind of usage iscertainly more useful in terms of user comfort It has been illustrated inscenario 2 (line 0 with more than one alert); animations are used during theopening and closing of a list. This could be generalized to every menu openingand closing.

CARE INO


Sound remains a very new media, difficult to use and to design. However,through the understanding brought by the state of the art review and the firstdesign experiments carried out during the project, it seems that there are realbenefits to be expected from a better management of sound in ATCenvironments. The benefits are to be expected in two directions:

An increased attention taken to sound issues when designing a controlroom, so as to avoid the clutter that is common in work rooms. Thehuman brain keeps trying to interpret the sounds it receives, and notbeing able to locate their origin, their cause or their meaning mightincrease the cognitive workload.

Designing sonic interfaces that carry appropriate information ornotifications. Because of its relations with attention, sound is very efficientfor notifications, and can also be designed to provide for less consciousfeedback. And as far as information is concerned, domains like the cinemaindustry have proven that sound can be purposely designed to convey thedesired information, even though the design dimensions are very differentfrom those of graphics.

6. Conclusions

During the first year of the project, the ANIMS team has focused on threeaxes of research: understanding the nature and uses of animation and soundthrough a state of the art review, identifying potential benefits and candidateapplications in air traffic control and demonstrating possible uses.

That approach, carried out with both classical research methods andmultidisciplinary design workshops, produced a wealth of information that canbe reused in the design of ATC systems.

First of all, the state of the art analysis of animation allowed the group toproposed definitions of animation that are operational for designers,programmers and human factors specialists; it produced an exhaustive list ofanimation techniques that designers can apply in user interfaces, a review ofthe most common uses of animation of user interfaces (notification, feedback,narration, etc), and a review of the human factors benefits documented in theresearch literature. It also provided some insight on the software techniquesused to produce animated systems. The same analysis on sound revealed afar less mature domain, and the work focused on understanding the nature ofsound perception, the reasons why sound created strong feeling among usersof interactive systems, and the techniques used in the entertainment industry(film, music, and games) to manage the media. More than a structured paletteof techniques, it produced a better understanding of the stakes of sounddesign, and research axes for producing efficient, non intrusive and meaningfulsound notifications.



Based on that knowledge, the work with ATC experts allowed the selectionof five ATC scenarios that are likely candidates to take advantage of the mostdocumented benefits of sound and animation. The scenarios includedistinguishing different alerts (STCA and down linked ACAS-RA), accessing tomore information in line 0 of a flight label, presenting STCAtype/urgency/importance, identification of the calling flight, animations andsound in a sequence manager

Finally, the design and prototyping phase produced demonstrators that inmost cases provide informal evidence of the benefits of animation, and in anycase provide insight as to the design issues attached to sound and animation.That phase, carried out using multidisciplinary iterative design techniques,revealed more practical issues related to the design, specification anddevelopment of animated and sound-enabled user interfaces: the need fortechniques to communicate and share opinions about animations and sounds,the need for techniques to specify animation and sound in an application, andthe need for techniques to integrate them in the architecture of an existinguser interface. That phase was also the opportunity for producing reusabledesigns (in the case of animation) and exploring design issues (in the case ofsound).

7. Perspectives

Considering the promising results gained during the first year, the perspectivescould be to further and depther evaluate the potential and applicationdomains of these animation and sound technologies applied for ATM. Thiscould include:

Possible work on the evaluation of the findings of the first phase: how canthe benefits of using animation and sound be proved empirically? Can theybe demonstrated or proved in more realistic ATC environments?

Work on supporting the design and development of animated userinterfaces. This more technical work would include research on theappropriate models, tools and architectures for describing animation,including them in user interfaces, and transferring them from designers toprogrammers.

Work on new ATC tools that would take advantage of the benefits ofATC and sound for close workers: by combining those benefits with thenew flat screen technologies and the new multiple-point input systems, itwould be possible to design interfaces on a touch-screen that would beshared by two controllers on a sector.

Work on creating reusable designs. The industrial success of graphical userinterfaces came not only with tools for producing them, but with good

CARE INO


visual designs that could be directly reused. Making animation morewidespread will probably require the same method; hence the need topropose reusable animated themes.

Work on designing sound for user interfaces. The first phase of theproject has provided insights and alleys of research for making sounddesign a more mature activity, comparable to graphics design for userinterfaces. Those insights now need to be used, and those alleys need tobe explored so as to propose design elements and methods.

Dissemination, training and creation of a design community. ANIMSprovides the basis for creating training sessions on the design of animateduser interfaces, creativity seminars, awareness seminars on the importanceof sound and graphics in ATC user interfaces. It also shows theimportance of creating a design community that brings together theappropriate knowledge and shares ideas in the same way that there areresearch communities in ATC or design communities in other domains.

All project information and report are available from the CARE INO web siteat the following URL:

http://www.eurocontrol.int/care-innov/public/standard_page/innov2_anims.html

Neural Network-Based Recognition and Diagnosis of Safety Critical Events


NEURAL NETWORK-BASED RECOGNITION

AND

DIAGNOSIS OF SAFETY-CRITICAL EVENTS

National Aerospace Laboratory NLR, The Netherlands

Foundation for Neural Networks, University of Nijmegen, The Netherlands

Supervised by Marc Bourgois.

1. Introduction

Successful safety management in air traffic management (ATM) needs an up-to-date picture of the safety of the operations. Currently, the most importantsource of feedback on trends in ATM safety levels is obtained from safetyoccurrence reporting by human operators, such as air traffic controllers andpilots. In an effort to support further development of ATM safetymanagement, research has been done in the CARE INO 2004 project on thefeasibility of a neural network-based system for automatic recognition anddiagnosis of non-nominal (potentially safety-critical) events in ATM.

2. Neural networks applied to ATM safety

Neural networks and related machine learning techniques provide thepossibility to learn associations between sets of signals. In the context of ATMsafety monitoring they may learn mappings between safety-relevant andobservable operational data, on the one hand, and the occurrence of aparticular type of safety event, on the other hand. The operational data thatmay be used depends on the operational context and may include, e.g., radartrack data, down-linked aircraft data, in-flight recorded data, air-groundmessages and ATC system input.

Development of a neural network-based detection system requires data forlearning of the associations. In this study, the suitability of several potentialdata sources for neural network-based safety monitoring was evaluated.These data sources included Airborne Collision Avoidance System ResolutionAdvisories (ACAS-RA’s) data gathered by Eurocontrol’s Automatic SafetyMonitoring Tool (ASMT), data of the human error database HERA-JANUS,ATM incident and operational data, and Monte Carlo simulation data for air

CARE INO


traffic operations. This evaluation considers aspects such as observability ofthe data types in an operational context, quantity of the data and types ofrelated safety occurrences:

ASMT ACAS-RA data: The advantage of using the ASMT ACAS-RA datafor neural network safety monitoring is that real operational ATM datacan be used for the classification of one or more issue categories ofdetected ACAS-RA events. In particular, the issue category ‘Level offabove / below’ is of interest for the ACAS-RA event investigators. Suchcategorisation is part of incident analysis, which was judged as animportant issue in automatic safety monitoring. About 500 to 600 eventsare expected to be available for which about 45% is expected to be in thecategory ‘Level off above /below’, as categorised by incident investigators.This amount of data is expected to be sufficient for neural networktraining, although practical application has to show whether thecomplexity of this classification task can be well represented by a neuralnetwork. In conclusion, this neural network classification task is a well-bounded problem with a reasonable amount of data, which makes it apotentially valuable application to research the feasibility of neuralnetworks within ATM safety monitoring. Section 6 presents the methodsand results of a neural network-based safety monitoring application basedon ASMT ACAS-RA data.

LVNL operational and incident data: The advantages of using LVNLoperational and incident data for neural network safety monitoring arethat real operational data can be used for neural network-based detectionof safety occurrences and for diagnosis of safety occurrences. Thisdiagnosis may be supported by severity classifications of safetyoccurrences as well as by safety assessments, which have been performedfor operational concepts at Schiphol Airport. As such, on the basis of theLVNL data neural network-based safety monitoring may be studied to abroader extent than is envisioned for the application with ASMT ACAS-RA data. The combination of operational data and safety occurrence datarepresents about one year and 2,000 safety occurrences. Application ofneural network-based safety monitoring on the basis of this LVNL datacombination would first require a proper choice of relevant safetyoccurrences and related traffic scenarios. It would be the subject offurther research to investigate whether sufficient data would be availableneural network-based detection and analysis for specific safetyoccurrences. In such further research it may be valuable to investigatecombination of the LVNL data with Monte Carlo simulation data and theresults of safety assessments. In conclusion, further research in co-operation with LVNL would be required before their data may be actually



applied for training of a neural network-based safety monitoring system.Given the attracting aspects of the LVNL data it may be worthwhilepursuing this further research.

Monte Carlo simulation data: The advantages of using Monte Carlosimulation data for neural network-based safety monitoring are the largeamounts of multi-source data and the clear coupling to accident riskassessment. Furthermore, Monte Carlo simulation data can be used toresearch under what conditions neural network-based recognition andanalysis of safety occurrences is possible. This knowledge may supportimplementation of neural network-based safety monitoring on the basis ofreal operational data. The performance on actual operational data thatmay be achieved by a neural network trained on Monte Carlo simulationdata depends on the inclusion of key aspects for recognition and diagnosisof safety occurrences in the model underlying the Monte Carlosimulations. A bias and uncertainty assessment, including a sensitivityanalysis, may shed light on the most important signals to support a goodperformance. The performance of a neural network-based safetymonitoring system may be supported by combination of Monte Carlosimulation data and real operational data. Such combination of datasources may, for instance, use Monte Carlo simulation data as training dataand operational data as test data to evaluate the capability of the trainedneural network. Another possibility for combination of these data sourcesis training of separate neural networks on the basis of various data sourcesand combination of the outputs of these networks in a HierarchicalMixture of Experts system. In conclusion, Monte Carlo simulation datamay support neural network-based safety monitoring by providing insightwhat conditions are required for such a system and by combination ofemulated and operational data. Further research into these issues wouldbe required.

3. Results

In this study the feasibility has been addressed of automatic detection anddiagnosis of safety occurrences by neural networks and related machinelearning approaches. Such a system may further enhance current automaticsafety data gathering in ATM. Neural networks and related machine learningtechniques offer the potential to learn mappings between a feature space anda class space. For air traffic safety management, the feature space representsair traffic characteristics (e.g., velocity or height at characteristic points) andthe class space represents types of safety occurrences.

For effective detection of safety-critical events by neural networks andmachine learning techniques progress has to be established on the following

CARE INO


areas. Suitable data sources on safety-critical events must be identified;suitable neural network / machine learning methods must be chosen, includingdata preprocessing methods. There must be sufficient data to train (optimise)the safety event detection/classification system. There must be additional data(not used for training) to validate the detection/classification performance.Each of these areas has been addressed in this research.

The identification of appropriate data is of crucial importance for the successof this approach. As such, the emphasis in this initial study was onidentification of requirements for data sources, identification of available datasources and subsequent evaluation of these data sources. Currently availabledata sources include Airborne Collision Avoidance System ResolutionAdvisories (ACAS-RA’s) data gathered by Eurocontrol’s Automatic SafetyMonitoring Tool (ASMT), data of the human error database HERA-JANUS,ATM incident and operational data of LVNL, data of the NLR Air SafetyDatabase, and TOPAZ Monte Carlo simulation data of NLR for accident riskassessment.

A first evaluation of these data sources showed, that ASMT data on ACAS-RA events, LVNL operational and safety occurrence data, and air trafficMonte Carlo simulation data are data sources to be further explored forneural network-based safety monitoring. In particular, these data sourcesinclude a range of ATM data, which may be directly measured in ATMoperations and serve as a basis for neural network input features, andknowledge is available about related safety occurrences.

For an initial ASMT-based data set of ACAS-RA events, the feasibility ofneural network-based learning was evaluated for automatic classification ofdetected ACAS-RA events. The classification problem concerned the ACAS-RA issues level off above/below and followed, which both could be classifiedas true or false. The data set contained flight-related data, such as track dataand RA data, on the one hand, and expert-based classifications of the ACAS-RA issues considered, on the other hand. The goal of this application was todemonstrate the feasibility of automatic system-based classification, based onlearning from the classification performance of a human expert. Theapplication included data pre-processing, training of several neural networkand machine learning techniques, and evaluation of the classificationperformance. Data pre-processing is of key importance for application ofmachine learning techniques. In the current application example, this consistedof selection of a limited number of static input features, which were believedto enable classification of the ACAS-RA issues (e.g., the vertical velocity of anaircraft at a number of instances around the time of an ACAS-RA event). Asa result, the automatic classification system in this example application used apart of the information that was available to the human expert.



For both ACAS-RA issues, the k-Nearest Neighbour classifier was found tobe the best performing machine learning approach with classification errorsthat could be as low as 18% of the events and up to a factor 2.1 smaller thanthe performance of a naive classifier. Thus, in spite of the limitation in inputinformation, learning from the human expert was accomplished. However,the achieved classification performance is not yet sufficient to actually supportthe safety analysis of ACAS-RA events. More advanced data pre-processingmethods, leading to more informative sets of input features and more insightin the expert labeling process are expected to further improve theclassification performance for this particular application.

Safety-critical events can be distinguished in consequences of safety-criticalsituations (e.g., separation infringement, ACAS-RA) and contributing causes(hazards) of safety-critical situations (Section 7). The example application ofSection 6 focuses on classification of consequences (ACAS-RA’s). Fordetection and classification of causes of safety-critical situations, one needsdata sets that incorporate these causes. Such early stage events provideinsight in the core of the development of safety-critical situations, do notneglect situations that may develop into safety-critical situations given theparticular circumstances, and can be naturally included in model-based riskassessment of current operations. Data sources, such as NLR’s TOPAZMonte Carlo simulation data for accident risk assessment and the identifiedLVNL operational data are of this type.

The training process of neural networks and machine learning techniquesrequires large amounts of data, especially for sophisticateddetection/classification problems with an extensive set of potentially relevantinput features. To acquire such large sets of data for causes of safety criticalsituations, using dedicated risk modeling-based Monte Carlo simulations is byfar the most feasible option. Moreover, Monte Carlo simulations can providemulti-source data from various agents in an ATM operation, which isimportant as basis for training of a system for detection and classification ofsafety-critical events related to multiple collaborating agents in air traffic.

For evaluation of the detection/classification performance of a system trainedon Monte Carlo simulation data, a separate data set that was not employed inthe training process should be used. Moreover, validation of thedetection/classification performance should be based on real ATMoperational data. For validation, the data quantity may be limited incomparison with the size of the training data set. The validation data mustinclude the sources to detect causes of safety-critical situations. A broad setof ATM operational data, with information from multiple agents in ATMoperations for a sufficiently large set of flights (such as, e.g., the ATMoperational data set of LVNL) is appropriate to this end.

CARE INO


4. Perspectives

Considering the results gained during one year, there would be a need tofurther research this innovative approach in ATM safety monitoring along thefollowing steps:

Identification of a suitable operational concept and associated hazards;

Development of ATM risk model and TOPAZ Monte Carlo simulator forthe concept identified;

Development of effective data pre-processing methods for neuralnetwork and rule based detection of causes of safety-critical events;

Training of the detection/classification system based on data setsgenerated by TOPAZ Monte Carlo simulations;

Validation of the trained system using sufficiently rich ATM operationaldata;

Evaluation of safety event detection system for ATM safety management.

These perspectives incorporate the recognition achieved in this researchconcerning the importance of data pre-processing methods that ensure thatthe automatic classification system has all relevant input information (e.g.,similar to all information used by a human expert in the ACAS-RAclassification example), the importance of detection of early stage events inthe development of safety-critical situations, and the identified availability ofTOPAZ simulation data for training and of ATM operational data forvalidation of the trained system. Development of such a safety monitoringsystem is expected to result in a valuable decision support tool for advancedATM safety management.


http://www.eurocontrol.int/care-innov/public/standard_page/innov2_neuralnet.html

The Airport of the Future : Central Link of Inter-Modal Transport ?


THE AIRPORT OF THE FUTURE

CENTRAL LINK OF INTER-MODAL TRANSPORT ?

M3 SYSTEMS, Space and Aeronautics SME, France.

ENAC, Aviation Economics and Econometrics Laboratory, Toulouse, France.

ANA, Aeroportos de Portugal S A, Portugal.


1. Introduction

In a context of a fast evolution of the air transport market, the future of theAir traffic Management will not only be linked to the improvements intechnologies, but also to the evolution of traffic flows. Despite the currentdifficulties in air transport, forecasts still mention strong traffic increases foryears to come. One of the main solutions chosen by the EuropeanCommission for coping with airport congestion problem and transports’pollution is to develop inter-modal transports to air. This development is animportant objective of the European Commission since Inter-modality andmultimodality are at the heart of the 2001 European Commission whitepapers on transport. One of the main priority objectives to be attained by2010 is to link-up transport modes for successful inter-modality.

2. What is inter-modality?

The first question what comes to mind is to know what inter-modality isexactly. What is its development today? More important what are theperspectives of inter-modality tomorrow in terms of airport development andwhat would be its influence on air traffic levels and distribution?

The study “The airport of the future: central link of inter-modal transport?”aims at providing answers to some of these questions when considering theglobal transport network. This constitutes an innovative aspect since theevolution of each transport mode was so far envisaged without takingnecessarily into account the evolution of the other modes, and ignoring thepossibility that the modes could be cooperative instead of being competitiveonly. An other innovative aspect of this study lies in the analysis of inter-modal transport as a way to tackle what could be the airport of the future; italso considers the inter-modality between all the possible transport modes.

CARE INO


Inter-modality is the characteristic of a transport network which allows theuse of at least two different coordinated transport modes for at least onesingle trip from origin destination. In literature, the term “inter-modal”transport applied to passengers using successively air and other transportmodes is used equally for the airport access to the city centre or for theintegration of the airport in the regional or national network of othertransport modes. As the implications of both types of airport inter-modalityare different in terms of investment, passenger needs, operators coordination,transport policies, etc., we have chosen in this study to differentiate betweenthem. In the case of airport access, the relevant modes to study are all publicmodes. In the case of integration of the airport in the regional or nationalnetwork, only rail is relevant (and particularly high speed train), since busservices on long distances are quite rare in Europe, and do not seem tobecome more prominent in the future. Conversely, air rail inter-modalityseems to offer promising opportunities for the future.

3. Key factor for inter-modality

The objective of the study is therefore to elaborate European scenarios oftransport network evolution by putting more focus on French and Portugueseones, and identifying the impacts of these scenarios in terms of developmentof inter-modality.

When studying what could be the role of inter-modal transport in the airportof the future, it is essential to determine what factors to be taken intoaccount in our analysis. The difficulty lies in the large number of factorsimpacting on the development of transport modes and in their complexrelationships. However among these factors it is important to differentiate thekey factors which are the basic factors influencing the transport demand andsupply (such as the world economy, the oil prices, etc.) from the resultingfactors which are the consequences of the key factors evolution (such as thelevel of traffic, of congestion, etc.). The relationships between these Key andResulting factors are used in the scenarios building



Oil prices World geopolitics

World economy

Passenger demand on leisure markets

Transport policies

Environmental and sustainable development concerns

TransportInfrastructure development

Operators’strategies

Multimodalcooperation

Multimodalcompetition

Group 1

Group 6

Group 4

Group 5

Group 3

Development of newTechnologies outside transport

Freight transport demand

Passenger demand on business markets

Group 2

Legend:

Transporttechnology

Key factor

Resulting factor

BChanges in factor A impact on factor B

Mobility

Unimodalcompetition

Traffic Congestion

Group 7

A

LEVEL OF AIRPORT INTERMODALITY

Relationships between Key and Resulting factors.

4. Inter-modality scenario

As baseline of our scenarios we consider that the evolution trends of some ofthe key factors will be the same for all of the studied scenarios. However, theextent of these trends can change between the scenarios. The association ofthe various nuances of these trends and of the key factors’ relationships hasled to consider three scenarios: a scenario A assuming a continuation in thecurrent instability situation, a scenario B assuming an evolution toward astrong instability situation and a scenario C considering a situation of globalstability. The main assumptions used in these scenarios are presented in Table1: Main key and resulting factors evolutions in all scenarios while the impacton the development of inter-modal agreements are detailed in Table 2:Scenarios’ results in terms of airport inter-modality.

CARE INO


Passenger demandScenario

Economicgrowthlevel

Environmental concerns

International tensions Oil prices

Business Leisure

Freightdemand

A1 Strong Moderate Moderateincrease

Highincrease

Moderateincrease

Highincrease

A

A2

High

Moderate Moderate Moderateincrease

Highincrease

Moderateincrease

Highincrease

B Low Weak High High increase Weakincrease

Weakincrease

Moderateincrease

C Moderate Strong Weak Weakincrease

Highincrease High increase Moderate

increase

Table 2: Main key and resulting factors evolutions in all scenarios.

Scenario

Level of useof air/HSTinter-modalagreements

on passengers’markets

Level of useof air/rail

inter-modalagreementsfor airport

access

Level of use ofair/bus

inter-modalagreements forairport access

Level of use ofair/rail

inter-modalagreements onfreight markets

Level of use of air/roadinter-modal

agreements on freightmarkets

A A1 Moderate Moderate Moderate High Moderate

A2 Weak Weak Moderate Moderate Moderate

B Weak Weak Weak Weak Weak

C High High Moderate High Moderate

Table 3: Scenarios’ results in terms of airport inter-modality.

Analysis of these scenarios tends to show that a good economic growth isnot sufficient for strongly developing airport inter-modality, especially air/railone for passengers and air/road one for freight. In particular, the levels ofenvironmental constraints play an important role in this development. Inaddition, the globalization process stimulates economic growth but may resultin unequal wealth distribution. This process leads to positive effects on freighttransport growth and multimodal cooperation. Its effect on multimodalcooperation for passenger transport depends also on other factors and variesaccording to the scenarios.

Concrete applications of these scenarios have been made on the case ofFrance and Portugal, which by their difference in the current inter-modality



development can be considered as representing the situation in “Core”European countries and less developed or new European countriesrespectively. Indeed Portugal does not have yet inter-modal infrastructure butplans to integrate Porto airport in the future high-speed rail network. Francewhich already has inter-modal infrastructure also plans to improve theintegration of airports in the high-speed rail network and the airport access bydedicated rail links.

Nevertheless, the applications of scenarios lead to the conclusion that despitethe difference of current state of inter-modality in both countries, buildingnew infrastructure could not be sufficient for developing airport inter-modality. If as a base for inter-modal development, inter-modal infrastructurehas to be built, the future of airport inter-modality should also be largelyimpacted by the market conditions (economic environment but alsocompetition levels on the transport market), as well as the air capacityconstraints and transport policies. The association of some conditions couldpromote the development of inter-modal agreements between transportoperators while other conditions could impede it.

The Air Traffic Management evolution for the next 15-20 years could be verysensitive to the development of such agreements. If there are so far fewexamples where airport inter-modality impacted air traffic, the number ofthese examples could increase with the level of airport inter-modality, and theair traffic level and distribution would then be affected more and more. Wecan indeed assume that a strong development of inter-modal agreementscould noticeably decrease air traffic on short and medium-haul. Change intraffic flows compared to the current situation, could be sizeable and involvedeep changes in their traffic flow management. This could help to alleviate aircongestion problems. We can then wonder on what conditions airport inter-modality can be a solution to air traffic congestion.

5. Perspectives

We propose to answer this essential question in performing a new study. Byshowing identifying factors directly or indirectly influencing the developmentof airport inter-modality and showing their complex relationships, the study“The airport of the future: Central link of inter-modal transport?” can indeedbe considered as the first step of a deeper analysis. The next step wouldconsist in analysing these factors deeper and determine what could be theconditions bringing about the development of airport inter-modality andunder which conditions airport inter-modality could lead to redistribution ofair traffic. This economic study would provide an economic analysis of themarket conditions impacting on the inter-modality development, in particularoperators’ strategies, unimodal and multimodal competitions. Economic

CARE INO


instruments (such as for instance the introduction of a Kerosene tax) andpolitical or administrative measures (such as for instance new distribution ofslots) favouring airport inter-modality would also be identified and analysed.Finally, all the conditions influencing airport inter-modality and their impactson air traffic would be analysed. As a result of these analyses, strategicguidelines for inter-modal development would be provided.

In parallel to this economic study we propose to develop the AIRMOD toolaiming at measuring the level of inter-modality at airports and simulating howthese changes impact airport catchments area. Airport inter-modalityindicators would be elaborated so as to provide a concrete measurement ofthe inter-modality level and computed for each considered airport. For agiven airport, current levels of indicators as well as airport catchments areawould be shown using a specific geographic map as web interface. TheAIRMOD tool would also allow to modify assumptions and indicators’ levelsso as to observe impacts on airport catchments area evolution.


http://www.eurocontrol.int/care-innov/public/standard_page/innov2_airport.html

CREA Training


CREA TRAINING

DeepBlue, research and consultancy SME, Italy.


1. Introduction

During the year 2000 and 2003, in the context of the first CARE INOprogramme, CREA! Project developped an inter-disciplinary approach todesign aiming at encouraging intersections between art, design and technologyborrowing from each discipline practices, methods and experience for thedefinition of innovative concepts. The application of the approach in complexdomains like ATM is innovative and challenging since the exploitation ofdesign practices and methods from disciplines different from engineering andhuman factors is largely unexplored. Innovation in such domain is constrainedby a problem solving view, neglecting other factors like aesthetic, affective,cultural and emotional aspects of human cognition. CREA! leverages creativityin action in artistic domains and offers a process model to integrate a varietyof suggestions into integrated solutions. Being impressed by the application ofthis method, Eurocontrol has decided to disseminate it as a way to favorinnovation within the ATM.

2. CREA! Method

CREA! is not a new methodology but a systematic design approach thatintegrates best practices, partly documented in literature but mostly comingfrom the long experience of the authors and the design team that supportedthe CREA! project; a sound theoretical background including activity theory,distributed cognition and cultural psychology, and well consolidated designapproach based on user-centred design and participatory design (King,Anderson 1995).

First of all, the design process is presented as a co-evolutionary process inwhich user studies , concept generat ion and technologydevelopment/benchmarking are carried out in parallel (divergence phase) andthen integrated (the convergence phase) in form of concept scenarios to feeduser-centred and participatory design sessions. The CREA! Design process isthe integration of two well-known design approaches, coming respectivelyfrom the industrial design and from the information systems design fields. Theindustrial design approach has its specificity in the fact that the design and

CARE INO


realization phases are continuously fed by a concept generation activity. In thisphase, often defined “simulate to stimulate”, the designers develop andvisualize scenarios of use, re-conceive the brief of the project, and specify thequalities and the attributes of the service. The concept generation phaseallows a constant flow of innovation into the design process, going beyondthe mere interpretation of user needs, to stimulate the demand of newfunctionalities that will transform the way in which the users see andunderstand their environment.

CREA Training


The participatory approach to system design has been developed inScandinavia as a means to achieve a close fitting between user needs andhabits and the new system to be developed through the direct participationof the users to the various phases of system development. In the last tenyears, participatory design has been enriched both because it has beencomplemented by ethnographic methods of analysis and because it has beenprovided with a new technology directly oriented to support the practices ofthe users rather than to automate parts of their work. For sake of clarity,usually participatory design approach misses the concept generation activity,and industrial design approach underestimates the social observation phase.

The approach intertwines the cycles in which the User Driven and the DesignDriven development are parallel, intersecting them frequently to compare theresults and re-tune the process. Representative users of the applicationscenarios will be directly involved in the process. This shall generate anextraordinary diversity of ‘components’, technologies and integratedprototypes. All these activities need be performed in strict collaboration withend users at different levels. Towards the end of the development, the twoparallel processes are supposed to merge and to produce one final result thatwill take advantage of both contribution and manage both innovation anduser adaptation.

The proposed approach suggests to adopt cheap, fast and easy to usemethods that still achieve the goal of designing effective solutions that affecthuman activity providing added value. Techniques include brainstorming, focusgroups, mock-up development, storyboarding, scenarios, walkthroughs andparticipatory heuristic evaluation. These techniques allow users to enhancetheir ability to relate to the design team without being afraid of the innovationprocess (which is usually associated with a loss of control), and ease thecreative process however innovative the concepts are. Rising above currentpractices requires a connection to them and the value of mock-ups,prototypes and scenarios stems from this view of epistemology. CREA!proposes a systematic evaluation along all the main phases of the designprocess: the project vision, high level concepts, activity models, conceptualrequirements, concept scenarios, early mock-ups and functional specifications,low-fidelity prototypes and interactive prototypes integrated in full scalescenarios.

3. CREA! training

The CREA! training was intended as an opportunity to investigate thepotentiality of developing innovative concepts for the Air Traffic Controldomain using a multidisciplinary approach based on the interviewing ofdifferent perspectives, from interaction concept creation and development, to

CARE INO


user involvement and early concept testing, to technology development andimplementation and on the activation of a design process based onnegotiation and meaning building and on the alternation of divergence andconvergence phases.

The training program was organised as an initial design experience aiming atdeveloping the following expertise/skills:

Activation in the participant, as integration of the typical “problem solving”approach, an attitude for “problem setting”, in which the operationalknowledge acquired is applied to an autonomous capacity for criticalelaboration;

Development of research skills, which round out the examination ofdesign themes;

Stimulation of the capacity to integrate the design choices involved in adesign solution or a service in a wider vision of the processes ofmanagement and in a strategic project involving all the material andimmaterial aspects of the offering (communication, service, corporateidentity, relations).

Engendering of an interactive design vision that accommodates within thecreative process an awareness of the practical implications of the choicesmade (design direction), crossing strategic aspects with specific domainconstrains.

Experimentation and understanding of the role of the directorship in amultidisciplinary process

Simulation of a design process reflecting on the selection, utilization andadaptation of the proposed techniques

Investigation of aspects related to the scalability of the CREA! approachand to its adaptability in relation to specific objectives and scale of theproject.

Exploration, expression and communication of new concepts and ideasthrough creative methods based on visual techniques like storyboardingand scenario

Critical reflection of the results of the workshops analysing the aspectsrelated to the scalability, adaptability and generalization of the envisionedsolutions.

The training was introductory to the methodology allowing participants toexperiment the proposed approach with the support of expert designers.The introductory module is oriented to give the participants an insight on theproposed approach introducing its key aspects from a theoretical andpractical point of view. It is articulated in the form of:

CREA Training


An introduction to the CREA! approach key aspects, rationale andtechniques.

A workshop on concept generation.

A workshop on Activity Analysis, Prototyping, Evaluation.

Participants were divided in groups involved in workshops on conceptgeneration and activity analysis. Each group was supported by a senior designand one or two experts in concept generation and/or activity analysis. Inparticular during the concept generation workshop participants were asked todevelop of a number of divergent concepts through group brainstorming; tocreate a mapping of the generated of concepts and then to refine thegenerated concepts. This enabled the participants to develop:

ethnographic observation;

participatory design: user involvement techniques;

activity modeling;

scenario creation; concepts, mock-up and prototypes evaluation.

All project information and report are available from the CARE INO web siteat the following URL

http://www.eurocontrol.int/care-innov/public/standard_page/project2002_crea.html


INOPublications

INO Publications


1. EEC Reports and Notes

[INO-04-01] L. Guichard, S. Guibert, H. Hering, J. Nobel, D. Dohy, J-Y. Grau,K. Belahcène 2004. Paradigm SHIFT Operational Concept Document V1.0EEC Note No. XXX/2004.

[INO-04-02] R. Ehrmanntraut, 2004, Total Information Sharing for PilotSituational Awareness Enhanced by Intelligent Systems (TALIS) - FinalReport, EEC Report Nr. 398/04

2. Conference Articles

[INO-04-03] J.Y. Grau, G. Gawinowski, L. Guichard, S. Guibert, J. Nobel, D.Dohy, & K. Belhacene, 2004. SuperSector Experimental Results: Proof ofConcept Assessment. Proceedings of the 23rd Digital AvionicsConference DASC , Salt Lake City, USA, Oct. 24-28.

[INO-04-04] R. Ehrmanntraut, 2004, Total Information Sharing for PilotSituational Awareness Enhanced by Intelligent Systems, in Proceedings ofthe 23rd Digital Avionics Conference DASC , Salt Lake City, USA, Oct.24-28.

[INO-04-05] R. Ehrmanntraut, 2004, Bandwidth Simulations of The TrafficInformation Service in Contract Mode (TIS-C) Over VDL Mode 2 With TheACTS Simulator, in Proceedings of the 23rd Digital Avionics ConferenceDASC , Salt Lake City, USA, Oct. 24-28.

[INO-04-06] R. Ehrmanntraut, R. Christien, 2004, Analysis of Aircraft ConflictGeometries in Europe, in Proceedings of the 23rd Digital AvionicsConference DASC, Oct. 24-28 2004, Salt Lake City, USA

[INO-04-07] R. Ehrmanntraut, 2004, The Potential of Speed Control, inProceedings of the 23rd Digital Avionics Conference DASC, Oct. 24-282004, Salt Lake City, USA

[INO-04-08] Marcus Lange, Jonas Hjalmarsson, Matthew Cooper, AndersYnnerman and Vu Duong - 3D Visualization and 3D and Voice Interactionin Air Traffic Management – In Proceedings of SIGRAD2003, The AnnualSIGRAD Conference. Special Theme – Real-Time Simulations,November 20–21, 2003, Umeå University, Umeå, Sweden andLinköping Electronic Conference Proceedings ISSN 1650-3686 (print),1650-3740 (www), February 2004, Noorkopings, Sweden.

[INO-04-09] M. Tavanti, T. Dang Nguyen, H. Le-Hong, 2004, UsabilityInspection of a 3D Interaction Metaphor. In Proceedings of the 2ndInternational Conference RIVF '04 - Research Informatics Vietnam-Francophony, Hanoi, Vietnam, February.

INO Publications


[INO-04-10] H. Hering, 2004, Wheelie - A mobile horizontal display filter toease controller’s separation task, In Proceedings of the 23rd DigitalAvionics Conference DASC , Salt Lake City, USA, Oct. 24-28.

[INO-04-11] H. Hering, 2004, Vital - an advanced time-based tool for thefuture 4D ATM environment, In Proceedings of the 23rd Digital AvionicsConference DASC , Salt Lake City, USA, Oct. 24-28.

[INO-04-12] J. Prinz, M. Sajatovic, H. Hering, 2004, S2EV-Analog Voice withenhanced Safety and Security, In Proceedings for 49th Air Traffic ControlAssociation (ATCA) conference, Washington DC, USA, November

[INO-04-13] C. Gwiggner, P. Baptiste, V. Duong. 2004 Some Spatio TemporalCharacteristics of the Planning Error in European ATFM. In Proceedings ofthe 7th International IEEE Conference on Intelligent TransportationSystems. ITSC 2004, Washington, D.C., USA, Oct. 3-6.

[INO-04-14] C. Gwiggner, G. Lanckriet. 2004, Characteristics Classes in FlightData - Estimation with Logistic Regression and Support Vector Machines. InProceedings of the 1st International Conference on Research in AirTransportation. ICRAT 2004, Zilina, Slovac Republik, November 22-24.

[INO-04-15] C. Gwiggner. 2004, Implicit Relations between Time Slots,Capacity and Real Demand in ATFM. Proceedings of the 23rd DigitalAvionics Systems Conference,

[INO-04-16] F. Ferchaud, V. Duong, C. Gavoille, M. Mosbah A New SlotAllocation for ATFM. In Proceedings of the 7th International IEEEConference on Intelligent Transportation Systems. ITSC 2004,Washington, D.C., USA, October 3-6.

[INO-04-17] F. Ferchaud, V. Duong, C. Gavoille, M. Mosbah 2004, UsingAbsorption Areas to Improve ATFM. In Proceedings of the 23rd DigitalAvionics Conference DASC , Salt Lake City, USA, Oct. 24-28.

[INO-04-18] F. Ferchaud, V.Duong, C. Gavoille, M. Mosbah. 2004. ReducingDisturbances by Using Absorption Areas. In Proceedings of the 1stInternational Conference on Research in Air Transportation. ICRAT2004, Zilina, Slovac Republik, November 22-24.

[INO-04-19] A. Cokasova, 2004. Air/Rail Intermodality Transport in Europefrom the Passenger Perspective, Dissertation Minimum University of Zilina,Slovakia, April.

[INO-04-20] A.Cokasova, 2004. Intermodality from the Passenger Perspective,In Proceedings of the 24th International Council of AeronauticalSciences, ICAS 2004 Yokohama, Japan October

[INO-04-21] A. Cokasova 2004. Passengers' Choice Between High-Speed Trainand Air Transport, In Proceedings of the 1st International Conference on

INO Publications


Research in Air Transportation. ICRAT 2004, Zilina, Slovac Republik,November 22-24.

[INO-04-22] P. Choroba, 2004. General approach for assessment of capacitybenefit by reduction of wake vortex separations, Perner's contact seminar,Pardubice, Czech Republic, February

[INO-04-23] P. Choroba, 2004. Airport capacity assessment with fast-timesimulation, University studies - Journal, University of Zilina, May

[INO-04-24] W. Glover and J. Lygeros, 2004. A Stochastic Hybrid Model forAir Traffic Simulation. in Hybrid Systems: Computation and Control(HSCC05), R. Alur and G. Pappas, eds., no. 2993 in LNCS, pp. 372-386,Springer Verlag,.

[INO-04-25] Nguyen-Duc M., Duong V., Drogoul A., 2004. Agent-basedmodeling and experimentation for Real-time Collaborative Decision Makingin Air Traffic Management, In Proceedings of the 24th InternationalCouncil of Aeronautical Sciences, ICAS 2004 Yokohama, Japan, October

[INO-04-26] Nguyen-Duc M., Boucher A., Drogoul A., Duong V., 2004.Towards Participatory Design of Agent-Based Simulations – An application inAir Traffic Management, In Proceedings of the 5th InternationalConference on Agent-Based Simulation, ABS 5, Lisbon, May.

[INO-04-27] Nguyen-Duc M., Duong V., Drogoul A., 2004. Conception d’unsimulateur multi-agent pour la gestion du trafic aérien, RechercheInformatique Vietnam & Francophonie 2004, RIVF’04, Hanoï, February.

[INO-04-28] Nguyen-Duc M., Duong V., Briot J. P., Drogoul A., 2003. Anapplication of Multi-Agent Coordination Techniques in Air TrafficManagement, International Conference on Intelligent Agent Technology2003, IAT’03, Halifax,

[INO-04-29] N. Archambault, G. Granger, and N. Durand, 2004, Heuristiquesd'ordonnancement pour une résolution embarquée de conflits aériens parune méthode séquentielle, CENA-INPT, in Actes de la DeuxiemeConference Internationale Associant Chercheurs Vietnamiens etFrancophones en Informatique (RIVF'04), Feb. 2-5 2004, Hanoï, Vietnam

[INO-04-30] D. Gianazza, N. Durand, and N. Archambault, 2004, Allocating3D trajectories to air traffic flows using A* and genetic algorithms, CENA-INPT, in Proceedings of the International Conference on ComputationalIntelligence for Modelling Control and Automation (CIMCA'2004), Jul.12-14, University of Canberra, Australia

[INO-04-31] N. Archambault, N. Durand, Scheduling heuristics for on boardsequential air conflict solving, CENA-INPT, In Proceedings of the 23rdDigital Avionics Conference DASC , Salt Lake City, USA, Oct. 24-28.

INO Publications


[INO-04-32] N. Archambault, Speed uncertainty and speed regulation inconflict detection and resolution in Air Traffic Control, CENA-INPT, InProceedings of the 1st International Conference on Research in AirTransportation. ICRAT 2004, Zilina, Slovac Republik, November 22-24.

[INO-04-33] T.Riviere, 2004, Redesign of the European Route Network forSector-Less, CENA-INPT, In Proceedings of the 23rd Digital AvionicsConference DASC , Salt Lake City, USA, Oct. 24-28.

[INO-04-34] T. Riviere, 2004, Generating a European Route Network forSector-Less , CENA-INPT, In Proceedings of the 1st InternationalConference on Research in Air Transportation. ICRAT 2004, Zilina,Slovac Republik, November 22-24.

3. PhD Dissertations

[INO-04-35] Huy-Hoang Nguyen, Coordination des Avions pour la Résolutionde Conflits: une Approche basée sur le Graphe PERT Disjonctif. PhDdissertation No. D1493, University of Technologies of Compiegne,January 2004.

[INO-04-36] Huy Tran-Dac, Sectorisation Contrainte de l’Espace Aérien. PhDdissertation No. D1503, University of Technologies of Compiegne,France, May 2004.

[INO-04-37] Monica Tavanti, On the Relative Utility of 3D Interfaces. PhDdissertation ISSN 0282-7492, ISBN 91-554-6102-6, Uppsala University,Sweden, December 2004.

4. Edition

[INO-04-38] Proceedings of the 1st International Conference on Research in AirTransportation, Vu Duong editor, ISBN 80-8070-196-2, 420 pages,Zilina (Slovakia), November 2004.

[INO-04-39] INNOVATIVE RESEARCH Activity Report 2003, Vu Duong editor,244 pages, printed by EUROCONTROL Logistics and Support Services,2004.

A C T I V I T Y R E P O R T

2 0 0 4

INNOVATIVE RESEARCH

E U R O C O N T R O LE X P E R I M E N T A L C E N T R E

© European Organisation for the Safety of Air Navigation

(EUROCONTROL)

This document is published by EUROCONTROL in the

interests of exchange of information. It may be copied in

whole or in part, providing that the copyright notice and

disclaimer are included. The information contained in

this document may not be modifi ed without prior written

per-mission from EUROCONTROL.

EUROCONTROL makes no warranty, either implied or

express, for the information contained in this document,

neither does it assume any legal liability or responsibility

for the accuracy, completeness or usefulness of this

information.

EURO

CO

NTR

OL

EXPE

RIM

ENTA

L C

ENTR

EIN

NO

VA

TIV

E R

ESEA

RC

H A

CTI

VIT

Y R

EPO

RT

20

04

RAP INO COUV REPRO.indd 1 13/07/2005, 14:27:26

INO 2004 - Eurocontrol

Documents