97-18

8/4/2019 97-18

http://slidepdf.com/reader/full/97-18 1/46

USAARL Report No. 97-18

Aircraft Multifunction Display and Control System s:

A New Q uantitative Hum an Factors Design

Method for Organizing Functions

and Display Contents

BY

Gregory Francis

Purdue University

and

Matthew J. Reardon

Aircrew Health and Performance Division

April 1997

Approv ed for public releese, distribution unlimited.

U.S. Army Aeromedical Research Laboratory

8/4/2019 97-18


Notice

Oualified reauesters

Qualified requestersmay obtain copies from the Defense Technical Information Center (DTIC),

Cameron Station, Alexandria, Virginia 223 14. Orderswill be expedited if placed through the librarian

or other person designated to request documents from DTIC.

Change of address

Organizations receiving reports from the U.S. Army Aeromedical ResearchLaboratory on automatic

mailing lists should confirm correct addresswhen correspondingabout laboratory reports.

Disposition

Destroy this document when it is no longer needed. Do not return it to the originator.

Disclaimer

The views, opinions, and/or findings contained in this report are those of the author(s) and should not

be construedas an official Department of the Army position, policy, or decision, unless so designated

by other official documentation. Citation of trade names n this report does not constitute an offkial

Department of the Army endorsementor approval of the use of such commercial items.

Human use

Human subjectsparticipated in these studies after giving their free and informed voluntary consent.

Investigators adhered to AR 70-25 and USAMRDC Reg 70-25 on Use of Volunteers in Research.

Reviewed:

WlWk

JEFFREY C. RABIN

LTC, MS

Director, Aircrew Health and

Performance Division

Released for publication:

eview Corm-i-&tee

Col&el, MC, k&S

Commanding

8/4/2019 97-18


REPORT DOCUMENTATION PAGEOMB No. 070+0188

lb .Unclassified

RESTRlCrmE M4RKlNGS

2a. SECURITY CLASSIFICATION AUlHORllY 3. DlSTRlBUTlON /AVAlLABlLllY OF REPORT

Approved for public release, distribution

2b.DECLASSlFlCATlON I DOWNGRADING SCHEDULE unlimited

4. PERFORMING ORGANlZATlON REPORT NUMBER(S) 5. MONITORING ORGANKATlON REPORT NUMBER(S)

USAARL Report No. 97-18

6a.NAME OF PERFORMING O RGANKATlON

U.S. Army Aeromedical8b. OFFICE SYMBOL 7a. NAME OF MONITORING ORGAN KATION

U.S. Army Medical Research and MaterielResearch Laboratory

6cADDRESS@ly,State,ndZ/PCode)

P.O. Box 620577

Command-

7b. ADDRESS (City, Sfate, and ZIP Code)

Fort DetrickFort Rucker, AL 36362-0577

IFrederick, MD 21702-5012

6a. NAME OF FUNDING I SPONSORlNG 6b. OFFICE SYMBOLORGANlZATlON (~applicable)

9. PROCUREMENT INSTRUMENT IDENTlFlCATlON NUMBER

I

6~. ADDRESS (City, State, andZiP code)10. SOURCt OF FUNDING NUMBtRS

PROGRAM

ELEMENT NO. ACCESSION

3M162787A879

11. TlTLE (h/udeSewlify-)

0) ,Aircraft multifunction display and control systems: A new quantitative human factorsdesign model for organizing functions and display contents

UnclassifiedSECURITY CLASSIFICATION OF THIS PAGE

I

I

.

.

I

.

.

.12. PERSONAL AUTHORIS)

.Gregory Francis and Matthew Reardon

13a. TYPE OF REPORT 13b. TIME COVERED

Final

16. SUPPLEMEHTAL NOTATION

14. DATE OF REPORT (Year, 15. PAGE COUNT&WI, Day)

17 . COSATl CODES 16. SUBJECT TERMS (Co&me on mwtse necessary and Ueniify by thck numbetjFIELD GROUP SUB-GROUP

01 02 Cockpit design, hierar:hy , multifunction displays,

24 07workload

19. ABSTRACT(Continueonnsmuseifneoessary andithtifybybincicnumbetj

The objectives of this study were to review the current state of aircraft multifunctiondisplay and control system (MF'DCS) design methods and develop a quantitative method of

designing MFDCSs that incorporate important human factors issues. Reports in the

literature indicate that MF'DCS design can influence flight performance. However, current

design methods rely primarily on the designer's intuition and experience. MFDCSs in

aircraft cockpits use computer-generated graphics and symbology that have integrated and

largely replaced the myrtad discrete electromechanical flight instruments found in older

aircraft. While much is known about the physical and visual properties of ME'DCSs, less

known about which human factors are important for their design and use. MFDCSs may resul

in greater workload if the distribution of virtual instruments, graphical and text data,

and control functions in an n-dimensional structure of display pages places excessive

cognitive and psychomotor demands on pilots during either routine or emergency situations.A quantitative method was developed, involving the derivation of a weighted sum of

separate cost functions, each of which incorporates the effects of an arbitrary number o

human factors and MF'DCS design guidelines. The method models, using a high level of

I

I

I

20.DlSTRlBUllON / AVAlLABILl-lY OF ABSTRACT 21. ABSTRACT SECURllY ClASSIFICATlON

q UNCLABSIFIEDAINLWTED 0 SAME AS RPT. [3 DTICUSERS Unclassified

ZZa. AME OF RESPONSlBLE INDM DUAL 22b. TELEPHONE (/n&de Ama code) 1 22~. OFFICE SYMBOL

Chief, Science Support Center 1 (334) 255-6907 1 MCMR-UAX-SI

ID Form 1473, JUN 88 pmkus-mobsokLa SECURITY CLASSlFlCATlON OF THIS PAG

Unclassified

8/4/2019 97-18


19 . Abstract (Continued)

abstraction, a pilot's search for specific information or functions amongalternative hierarchies of MF'DC Sdisplay pages. An annealing algorithm wasproposed as an effective nume rical method for finding the display pagehierarchy that minimizes the compo site cost function. Further research isneeded to determine whether the set of constituent cost functions issufficient or needs to be expande d. Studies also are needed to determine

specific values for cost function coefficients and to validate the overallmodel. The quantitative method delineated in this report for designingoptimal hierarchies of ME'DC Scontent pages and functions may become usefulfor engineers as a design tool'during developm ent of MFD CSs that willmaxim ize pilot performance and minimize errors and excessive in-flightworkload.

8/4/2019 97-18


Table of contents

Page

Introduction .................................................................. .

Reducing information workload in the cockpit ...................................... .7

Integration ................................................................. .

HUDS .................................................................. ...8

Pilot’s associate ............................................................ .8

Alternative MFDCS interfaces ................................................ .9

Expanded use of visual and auditory senses ...................................... .9

MFDCS content and interface design ............................................. 10

MFDCS design issues ....................................................... 10

Quantitative MFDCS design methods ............................................ 14

Current state of MFDCS design ................................................ 19

A new quantitative method for optimizing MFDCS content hierarchies ................. .20

Cost as expected access ime ................................................. .20

Hill-climbing ............................................................ ..2 2

Cost for related functions .................................................... .23

Cost for expected access ime and relatedness ................................... .26

Simulatedannealing.........................................................2 7

Costfunctions ........................................................... ..3 1

Frequentlyusedfimctions ................................................ ..3 1

Timecriticalfunctions ................................................... ..3 1

Ideallocations ......................................................... ..3 1

Repeated selection of buttons .............................................. .32

Minimize number of levels ................................................ .33

Minimize overall access ime ............................................... .33

Related functions on close pages ............................................ .33

Consistent ocation of related items .......................................... .33

. . .11 1

8/4/2019 97-18


Table of c ontents continued)

Page

Related unctionson the samepage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .34

Errors ................................................................ ..3 4

Dedicated displays ..................................................... ..3 4

Discussion ............................................................... ..3 5

Conclusions .............................................................. ..3 6

Refere nces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ...37

List of figures

1. AschematicoftheaftcockpitintheAH-lWSuperCobra,Venom .. .................. .3

2. Simulated ages or the proposed enom Super-Cockpit........................... .4

3. The designprocessor the development f the V- 22 Osp reycockpitand MFDCS ....... 12

4. A hierarchialstructu re ith three evels and threepossible ptions t eachchoicepoint ... 16

5. The development f a hierarchy hroughhi&climbing ............................. .24

6. The development f a hierarchy hat minimizesdistance etween elated unctions ..... .25

7. Hierarchies or minimization of ex pected ccessime andrelatedness ............... .26

8. An intuitive description f hill-climbing and simulated nnealing ................... .28

9. Freque ncy f final hierarchycosts or hill-climbing and simulated nnealing .......... .30

iv

8/4/2019 97-18


Introduction

Everyone, at least occasionally,has probably experienced rustrationwith poorly designed

interactive,computer-based,nformation display interfaces. Modem automated elephony

systems, or example, have had notoriously problematicuser nterfacedesigns. Users often have

had to listen to lengthy, complicated nstructionsand navigate heir way through numerous evels of

option menus to eventually reach or input the information they desired. Likewise, the typically

poor interfacedesigns or many bank automatic eller machines ATMs) have prevented ndividuals

from learning to use them (Rogers et al, 1996). This has frustratednew ATM usersand,

undoubtedly,has been costly to providers of these ypes of servicessince hey probably lost a

portion of these customers o competitorsoffering alternative,easier o use systems. In aviation

contexts, he quality of an interfacedesign for electronicdisplay and control systemscan obviously

have greater mpact than mere inconvenienceand frustration.

Military and civilian aircraft designed n the 1960’s and 1970’s had so many separategauges,dials, ights, switches,buttons,circuit breakers,control wheels, and levers n compact aircraft

cockpits hat crewmembersnecessarilyhad to spenda significantamount of time heads-down

scanning nstrumentpanels o find the information and functionsrequired o maintain safe flight.

At that time, display, monitoring, and control functionswere still largely dependenton the use of

loosely nterconnectedanalog systems. With such echnology,maintaining continuous,complete,

and accurateawarenessof aircraft status mposed a heavy psychomotorworkload. It required

explicit mental effort to continuously ntegrate he dynamic information from the many scattered

dials, gauges,and advisory or caution ights. Furthermore,an early or subtleemergency situation

probably ook longer to clearly identify, and more steps o correct, han is usually the case n

currentgenerationaircraft. Becauseof high pilot workloadsassociatedwith early generation

cockpits,most transportaircraft required a flight engineer n addition to the pilot and copilot.

The development of increasinglycapablemicrocomputers,software ools for implementing real-

time digital data acquisitionsystems,and advances n the design and manufactureof small video

displaysprovided the technology for the evolution of computerizedmultifunction display and

controlunits for both military and civilian aircraft. Technological advancesgradually permitted

replacing he multitude of separate lectromechanical tatus,warning, and control deviceswith

integratedmultifunction display control systems MFDCSs). From their inception, MFDCSs were

often similar in appearance nd usage o ATMs in that crewmemberspushedbuttons o move

througha hierarchy of display pagescontaining nstructions, nformation, or lists of user-activated

functions e.g., data entry). MFDCSs gained increasingacceptance mong aviatorsand were

generallycreditedwith reducing cockpit instrument“clutter” as well as reducing the time

crewmembersspentsearching or, and mentally integratingaircraft status nformation. The

reduction n pilot workload due to the introductionof increasinglycapableMFDCSs in the cockpit

was a primary factor in eliminating the need for flight engineers n most currentgeneration

transportaircraft.

1

8/4/2019 97-18


The initial impressionsf M FDC Ss were hat they reduced ilot workloadduring outine light.

However,with time, any reductionsn w orkloadweregradually ffsetby the ability of these

computer-basedockpitsystemso encapsulaten increasing umber f additional eatures,

functions, ndcapabilities ot feasiblewith the analogsystemshey replaced.This progressive

increasen functionality asbecome articularly pparentn m ilitary aircraft. For exam ple,

military com bat ndelectronicwarfareaircrafthavebeenusingcomputer-basedisplaysystemssince he 1970 ’s and ,although oday’s versions f thesesystems avemuchgreater omputational

speeds an dmemorycapacity, he num berof functions vailable o users e em so haveexpandedproportionately.Most of the expanding rrayof functions equire ubstantial rewm ember

involvement e.g.,monitoring largeamountof ad ditional, reviously navailable, nformation;

selectingrom an expanded rrayof options ndsystem onfigurations; ultisensor-basedecision

making;and roubleshootingomplexsoftware-dominatedystems). herefore, rewmem ber

workloadswith current tate-of-the-artircraftMFD CS s n some ircumstances ay actuallybe

greaterhan hatexperiencedn olderaircraftwith lesssophisticatedystems.

The useof MFDC Ss n U.S. Army helicopterss ust beginningo become revalent. Currently,for example,only the OH -58D scout elicopters ndversions f the UH -60 utility transportorspecial perations avemore hanoneMF DC S in the cockpit nstrument anel. OtherArmy

helicopters reprimarilyequipp ed ith the more raditional rrays f discrete lectromechanical

gauges, ialsandswitches. How ever,helicopter pgrades ndentirelynew helicopterdesignsor

the U.S. Army, such s he Comanche cout/attacknd he TiltRotor ransport elicopters, ill

includemultiple,highly integrated ockpitMFD CSs and etainonly a few critical backu p nalog

gaugeso ma intainbasic light capability n caseof complete lectronic ystemsailure.

Figure1 is a schem atic f the aft (copilot/gunner) ockpit ayoutof the AH-1 W SuperCo bra

attackhelicopter sproposedor the BritishArmy (Holley andBusbridge, 995). This is a modern

versionof the AH-l Cobragunsh ip, hich originallywasdesignedor, and effectivelyused n theViet Nam W ar. SuperCo bra rototypesncorporate n ad vancedechnologymissionequipment

package alled he SuperCockpit hich ncludes wo largecolorMFDCSswith 26 push-buttons

integratednto the surrounding evels. Eight of the push-buttonsrehard-keyswitches hich

activate riticalor frequently sedhigh-level unctions r displaymodes. The other 18 push-

buttons resoft-keys,mean ing hat heir functions nd abelsmay change cross ifferentMFDCS

displaypages.

Figure2 depictswo MFDC S displaypages or the SuperCockpitHolley andBusbridge, 995).The left displayshows eal-timestatusnformation rom the aircti engines nd otheraircraft

systemsSYS ). The push-buttonsn the right sideof the panelareassoc iated ith software-

generated isplay abels ndicatingumps o additionaldisplaypages ontaining elatedinformation.Pressing soft-keycauseshe MF DC S to displaya new pagecontaining he

information r functionsndicated y the key’s label.

MF DC Ss typically containa w ide rangeof singleandmultistep unctions.The type of objectsand nformationdisplayed n the MFD CS , the dataacquisition hannelshat are representedy the

2

8/4/2019 97-18


0MD

1 /@-

I STOWAGE I1_(16’x5.75’x-6’~

Figure 1. A schematicof the aft cockpit layout in the AH- 1W SuperCobra,Venom. TwoMFDCSs display the bulk the information.

3

8/4/2019 97-18


Figure2. Simulated agesor the proposed enom SuperCo ckpit. he systems age top left)

showsnformation n engines nd ncludes egends long he right o indicate hat pressinghe

associateduttonwill causehe display o pres enthe requestednformation.Targeting

information s show n n the top right figure. The hierarchical tructure orrespondingo someof

the MFD CS is presentedt the bottom.

4

8/4/2019 97-18


displayed bjects,he setof activedatabaseinks,aswell as he functionshat soft-keys anactivate

arecommonly roupedogetherogicallyon oneor more nterconnectedisplaypages orminga

specificMF DC S mode. Flight crews ancycle hrough he numerous FDC S functionalmodes

with oneor moreof the surroundingush-buttons.ypical MF DCS modesnclude hose or

attitude eference ndnavigation, omm unications, ovingmapdisplay,systems ontroland

status,argeting ndweapons election ndstatus, swell assituational warenessisplays asedon mu ltisensor atafusion. omeMFDCS modesmay havedisplaypage s ontaining lusters f

related irtual nstrumentsuch sattitude, ltitude, ndairspeedndicators,uel gauges,moving

maps, tc.,with or withoutsymbology verlays or navigation r weapons election nd argeting.

Displayscreensredesignedo present,or any selectedmode,only a subset f the total

information rom the monitored ircraftsystems. ilotsdynamically elect isplaymodes ased n

the nformation nd unctionality esiredo accomplish onstantly hanginglight managem ent r

combat asks uch ssituational warene ss,avigation, omm unications,ystems onitoring,battlefield nd h reatmon itoring, nd argeting.

An M PDCS canbe conceptualizedsa relativelysmall wo-dimensional indow or viewinga

singlepageof information electedrom a much argernumber f pages f staticanddynamicdata

arrangedn a mu ltidimensionalierarchy.The information ccessibleia an MF DC S and ts hard-

or soft-keyoptionselection uttons asa virtualstructurehatcanbe representedescriptively,

graphically, ymbolically, r asmathematical odels. For crewmem berso efficientlyusecomplex

andextensiveMFDCS dataand unctionhierarchies,hey mustacquire n accuratemental mage

andconceptual nderstandingf how all the dataand unctions ncapsulatedn the available

displaymodes regrouped nd nterrelated ndhow this structure anbe efficientlyand apidlytraversed sing he available edicated ndsoftware efinedbuttons. f the displaypagehierarchy

andnavigable aths etweenunctionally elated lusters f displaypages renot well understood,

MF DC S users re ikely to become ost n the MF DC S’s information pace r become onfused

with regard o the ocation f immediately eedednformationor functions.

Obviously, ecomingost n the information pace f a poorlydesigned FD CS wouldonly addto a pilot’s sense f da nger nd confusion uring n-flight eme rgenciesnvolvingspatial

disorientation,erious ystemailures,or sudde n nusual ttitudes.Duringcritical n-flight

situations herecomposure,larity of thought, ndefficientuseof time areessen tial, etting lost”

in the pagespace f an MFDCS is likely to precipitate anicandprevent dentification nd

resolution f the problem. n suchsituations, F DC S usersmightbeginentering ssentially

randomMFDCS pagenavigation elections. imilarly,duringcombat perations, unnersn

Army attackor scout elicopters, espite anger nd ear,mustbe able o rapidlyandaccurately

traversehe nformation MF DCS mode)subspaceelevant o their specializedasks.Gunnersn

high threatscenarios ustbe able o cyclevery rapidly h rough ariousMFDCS modes oaccomplish uch asks s argetdetection,ecognition, and-off, anging, rioriti&ion, w eapon

selection,argetdesignatione.g. asing),weaponsiring, andeffectassessment.ecoming

confused t any pointduring hese omplex rocesses,ith respecto how to transfer etweenmodes n the MFDC Ssutilized o perform he asks, ould esult n target scape r, of more

5

8/4/2019 97-18


immediate onsequence,ive the adversa ry ufficient ime to detec t, lose n, and ire first, with

potentially ethal effect.

ModernMF DC Ss are ruly impressive ndseem unctionally nd estheticallywell designed s

depictedn advertisementsnd duringdemonstrationsn circumstancesf little or no stres s.But,

while the modemcockpit elieson MF DC Ss, ittle hasbeenpublishedegarding ow unusual,critical,or dangerousircumstancesffectuser-MFDCSperformance ndmission ffectiveness.

Furthermore,hereha snot yet beena systematic valuation f M FDC Ss to enum erate nddefinea

taxonomyof the cognitivean dpsychomotoruman actor ssueshat should e considered uring

their design . n this report,we offer whatwe believe o be a new quantitativemethod or designing

MFD CS displaypagehierarchiesha toptimizeshe distribution f content nd unctions s inga

se tof weigh ted riorities epresen tinguman actors nddesign uidelineshought o be important

influencers f use r-MFD CS nteractions.

MFD CSs rade he workloadassociate d ith visually searching ockpit nstru men tsor a

cognitiveworkloadassociated ith a cognitivesearchhroughmental mages f a multi-

dimensional atabas e f pag es f information nd unctions.Physicallysearchingor a display

pagecontaining ecessaryunctions anbe time consuming ndoftenhas he additionaldrawback

of requiring he coordinated s eof button s, ursor ontrols, nddataentrykeypads.These

activities a ndistract rewmembersnd emp orarily educe heir situational ware ness . he

SuperCobrand he AH-64D LongbowApachecockpit includenumerous F DCS modeselect

buttons ndmenuscroll oggles ocated ot only along he borders f the MF DC S, b ut alsoon the

flight controls Hanne nand Cloud, 1995). Studiesndicate hat ime spent ccessingnformation

from a MFD CS influences erformance.Sirevaag , t al. (1993) had five U.S. Army helicopterpilots ly sim ulated ap-of-the-earthNOE ) reconnaissanceissions nd eport nformation t

specificwaypoints.Reporting his nformation equired aging hrough n MFD CS . Although he

pilotsalsohad a head-up isplay H UD) on their helmet hatprovided ircraftsituational warenessinformation spee d, ltitude,etc.), light perform ance asadversely ffected s he comm unication

load ncreased.n particular, nderhigh com municationoads, ilotsspent, n average, more

seconds erminuteabove h e specified OE altitude. That study llustratedhat the time spent

accessingnformation rom MF DC Ss canadversely ffect light perform ance.

Such indingsare consistent ith concerns bout he workload equiredn continuously

balancing light an daircraftsystems-managementuties.The capabilities f an increasing umb er

of aircraft equirecarefulattention o, andskilleduseof, manyMFDCSs. For example,Dohme

(1995) observedhat OH -58D A eroscout ndAH-64 Attackhelicopters othuse he airborne argethand-offsystem ATH S), accessedhrough n MF DC S unit. The databaseor the ATH S functions

aloneconsists f approxim ately 80differentpages f menus,nput ields,and nformation ATHSis alsoone of the options n figure2, top right). Dohmeestimatedha tabout300 pages f

information upportedhe entirese tof functionsn the MFDC S. He suggestedhat learningall the

MFD CS modes nd developinghe ability to quickly andefficiently a cces she relevant nformation

for all potential askswasa formidab le hallenge or trainees.

6

8/4/2019 97-18


Thereare concerns bout xcessive ircrewworkload dversely ffecting light performa nce

during omplicated r stressful issions.Duringhigh workloadmission egments,rewmembers

may begin o selectivelygnoreelements f informationwhichmay actuallybe quite m portant.

The nextsection iscusse s ethods f improving verall nformation cquisitionn the cockpit, nd

then ocuses n how to incorporate ognitive ndpsychomotoruman actor ssues, swell as

design uidelines,nto the MF DC S design rocess.Subsequentections ropose new method orincluding uman actor ssuesn determ ining n optimaldistribution f MFD CS content nd

functions, iscus s ow to apply he quantitativemethods, nd ecomm end irectionsor further

research.

Reducing nformationworkload n the cocknit

As military aircraftcomplexity nd unctional apabilitiesncreased,oncern rose hat

crewmembersouldbecomemoreeasilyoverwhelmed ith information nd askoverload. n

responseo this concern,heredeveloped strongnterestn simplifyingcockpitsystem-userinterfaces nd assisting ilots n copingwith theproliferation f flight andmission elated

functions.A general oal or new aircraftdesigns as o make t aseasyaspossible or crew -

memberso access, nderstand,ndefficiently akeactionon cockpit ndsystems-relatedata.

This section eviews arious ropose d ethodsor improving nformationransfer o crew-

memberso improve light andmission erformancendcapabilities.

Integration

The ntroduction f computer-drivenisplayandcontrolsystemsnto aircraftcockpits llowedMFD CS designerso create ew anddynamicmethods f combining ndpresentingnformation

from systems n dsensors. singlecockpitdisplaybecam e apable f simultaneouslyntegratingmanydifferentsource s f information, hus educinghe workload equiredo scana multitude f

separatenstruments.Work in this area ed to novelme thods f integrating ndportraying light

information reviewed y Stokes ndWickens,1988). In support f these fforts,a wide varietyof

new symbologywasdeveloped, ut often t wa sonly applicableo specific ircraft e.g.,Newman,

1995,Appendix).

Integratingnformation rom m ultiplesourcesnto an MFD CS cangreatly e duce he time

neededor crewmemberso accessnformation.Additional mprovem ent ouldbe gainedby

refining he criteria or selecting hichdisplayobjects nd soft-key unctions houldbe co llated

togethernto functionally elatedgroups f displaypages.The proper trategyn designinghe

contents, enus, ndbranching chem eor M FDC S page s as he potential or reducing he total

num ber f displaypages r modes.Com bining elated nformation nd unctionality nto

relatively ew coherent isplaymodes angive crewm embersbetterunderstandingf the entireinformation tructure ndallow fasterandmoreefficientus eof M JTDC S apabilities.As a result,automatedlight systemsnformationntegration anachieve argesavingsn creweffort.

7

8/4/2019 97-18


On the otherhand, ntegrating nrelatednformation ources ndsoft-key unctionsnto single

displaypages anhinder, a ther hanhelp,crew me mb er’s nderstand ingf systems tatusStokes

an dWickens,1988). Likewise, ncluding n excessive umberof menuoptionsor soft-key

functionsn a singledisplaypageor MPDCS modecanproduce isplayclutteran dcomplicate

crew me mb er’s earchor a particularunction. Decidingwhich unctions nd nformationobjects

to integrateogethernto a singleor relatedgroupof displaypages equires thoroughunderstandingf the interrelationshipetween ircraftsystems ndsubsystems,swell as he

information n d unctions equired or perform ing ockp itproceduresndmission asks. However,

dataand unction ntegration ased n these actors loneusuallywill n ot solveall MFD CS-user

interface roblem s.The MF DC S content atabase u stalsobe designedo incorporate isplay

pagesn a wa y that max imizes h e user’sability to efficiently search nd ocate he desiredMPDCS

functions, ptions, ndpages r modes.

HUDS

HU Ds project via application f advanced ideo echnologies)light informationdirectly nto

the crew perso n’sine of s ight, hereby educinghe need or head-down canning f cockpitpanel

displays r instruments.HU D systems llow pilots o continuouslyrackrelevant light

performance arametersia computer enerated ymbology nddatasuperimposedn the directline-of-sightmagery. Num erous tudies avedemonstratedmproved hght performance ith

HUD s (seeNewm an 1995 for a comprehensiveeview). Currently, owever,HUD s cannot

displayasmuchor aswide a rangeof different ypesof dataan ddisplayobjects s MFD CSs . This

is partly because xcessivenformation r displayobjects rojected n a HU D can ead o severe

visualclutter, herebydeteriorating pilot’s external iew. Therefore,HUD s do not supersedehe

need or MFD CSs . Increasingly ophisticated PDC Ss will continueo be the primary light and

systems onitoring ndmanagementnterface or civilian an dmilitary pilots or many decadesnto

the future. HUD and MFDC Ss, however,will undoubtedly ecom encreasinglyntegrated ndcomplementary.

Pilot’s asso ciate

A p ilot’s assoc iates an advanced onceptor assisting ilotswith a software-basedystemhat

uses ata usion echniques ndautomatically nalyzes om plexmultisensor ata, ecom mends

actions, nd mplem ents ilot’s com man dso performcertain asks.Partof the pilot-associate

interfacewill con sist f an advan ced ighly integrated PD CS utilizinga large latpanelscreen s

partof the user nterface. t w ill incorpo ratertificial intelligencemethodso ad aptively ntegratemultisensornformationan ddynamically dvise ndalert crews boutpotentialproblems,

solutions,hreats, nd opportunitiesMcBryanan dHall, 1 995). It will alsobe capable fautonomousecisionmaking or constrainedndpredefined ircumstances.he pilot’s asso ciate

will autom aticallyrackan danticipate ecessaryhangesn flight mo des nd adaptivelyorganize

an ddisplay he approp riateask-orientednformation nd unctions.The development f such

system as he potential o greatly educ ehe need or pilots o searchor and ntegratenformation

and unctions cattered mong he manydisplaypages r modesn an MFDCS.

8

8/4/2019 97-18


While potentially valuable, a pilot’s associate or advancedmilitary rotary-wing aircraft is still an

emerging echnology. Moreover, similar but lesscomplex types of automation n commercial

aircrafthave occasionally ed to seriousproblemswith “mode awareness,”whereby crews have

experienceddifficulty dete mining what the automationwas doing (Sarterand Woods, 1995).

Currently, mode identification often requirespaging up and down throughdifferent layers of the

MFDCS modes to enable he user o identify the most current settings or systemand control

variablesas well as to reorient with respect o location n the MFDCS mode or page hierarchy.

Alternative MFDCS interfaces

Researchsuggestshat using an MFDCS function select nterfaceother than push-buttonscan

reduce he difficulty of navigating hrough MFDCSs’ information and function space. Speech

recognitiondevicesand pilot electroencephalograghicEEG) signalsare potential means of hands-

off interfacingwith an MFDCS. Such methodseventually could replaceor complement the use of

hard and soft-keys for controlling MFDCS displays,selectingmodes, and activating various

functions. These alternative nput interfaceswould have the advantageof freeing the pilots’ handsfor other asks. However, they will not necessarily ead to improved performancesearchingan

MFDCS database.Reising and Curry (1987) found no difference in flight performance or a

speech ecognition nterfacecompared o a well-designedpush-button nterface. Whatever the

interface, imitations in the design of the MFDCS still will likely impact a flight crew’s ability to

lily exploit the many complex capabilitiesof the aircraft. Indeed, it may be necessary o entirely

restructurehe MFDCS databaseo obtain optimal performancewith a new interfacemethod. How

to do this rationally is not clear and requiresadditional MFDCS human factors esearch.

Expanded use of visual and auditory senses

Another alternative o the MFDCS interface s presenting light and aircraft systems nformationto crewmembers hrough peripheralrather than foveal vision. Stokesand Wickens (1988) provide

a review of studies hat evaluatedauditory and peripheralvisual displays. Information delivered via

a peripheraldisplay is designed o be noticeable n the pilots’ peripheralvision. Although

potentially useful, the benefitsof such displayshave not yet been verified in aircraft. Additional

researchs needed o define how they can be effectively adapted o enhancepilot performance,

information processing,situationalawareness, nd decision making.

Simple auditory signalsare commonly incorporated nto cockpit warning systems. However,

more complex warning and advisory auditory systems, o include three-dimensionalauditory

“displays” to assistcrews with situationalawareness nd threat ocalization,are being researched.

Major drawbacks or extensiveuse of auditory systemsare their potential or interfering with crew

communication, he time needed or listening to and interpreting ong messages, nd their transient

nature,which may requirepilots to rapidly refocusattention from other tasks o mentally register

the auditory message. These are some reasonswhy auditory pilot information systemsare unlikely

to completely supersede isually oriented MFDCS panels.

9

8/4/2019 97-18


MF DCS content nd nterfacedesign

MFD CS systems re typically composed f hardw are nd software om ponents. he hardwarecomponentsncludeaviationcapable om puter oards, ockpitdisplaypanels,surroundingevels

with push-buttons,ndalphanumericeypads.Software omponentsnclude eal-timeoperatingsystems,outines or generating ynamicsymbology,map databases,ircraf3 ystemsiormation,aswell asdatabasesor graphic isplayobjects, o ft-key unctionmappings, bject nteraction

rules,performanceimits, p rocedures,ndvarious hecklists.A governing vent-oriented oftware

program eeps he system ontinuously ctiveand esponsiveo pilot inputsand changesn aircraft

status.The design f this software, nd he databaseshat t candynamicallyand adaptivelydraw

objects nd nformation rom, s the focusof our concern. MFDC S software ndassociated

databasesanbe conceptualizedsa multi-dimensionalpace f interconnectedages f

information,menus, nd unctions.The high-leveldesign roblem s how to organize n optimal

structure ndpatternof interconnectionsor the nformation nd unctionalityassignedo an

MF DCS . Because f the complexity f these ystems,t is usuallynecessaryo defineoptimal&y

with respecto constraintsnddesired erformanceriteriaor goals.One of the essential esigngoals or an MF DC S is that users e able o efficientlysearchhrough ts informationspaceo find

necessary ataandcontrol unctionsn urgent ituations.

Carefuldesign nddistribution f displayobjects, ata,and unctions crossMFDCS pages ndmodes anmimmize he time an deffort requiredo locatenecessarynformation. For exam ple,

Reisingan dCurry (1987) used realisticF-l 5 simulator a mewhichprojected he out-the-windowview on a display n a cockp itmockup nd equired onpilot estsubjectso accesslight,

navigational, ndsystemsnformation hrough simulatedMFDC S. They comparedlight

performanceor two hierarchical esigns f the MF DC S displaypages.They foundsubstantial

improvementn flight performance hen hey organizedhe contents f the pages ccordingo the

differentphas es f the flights,comparedo a fixed organizationha tclusteredhe informationaccordingo datasource haracteristics.heir resultsndicated hat different ypesof MFDCS pagehierarchies ouldsignificantly nfluence imulatedlight performance.

Assigning imctions o pages ndswitchess a d ifficult taskbecausehe human-computer

interactionsnvolved n accessingnformation rom an MF DC S arecomplicated ndnot entirelyunderstood. nfortunately,he frequency ndpattern f MFD CS modeor pageswitching nd

functionselections uringactual light havenot beenwell docum ented. lso, the largenumberof

possible ombinations f pages,unctions, ndsoft-switchesuickly eads o combinatorial

explosionwhenattemptingo consider ll possibleayouts.The nextsectiondescribes urrentapproacheso MFDCS design.

MFDC S design ssues

Studies f human-computernteraction ave nvestigated any mportant ha racteristicsf

displays ndhum an nformation rocess ing. he displaysmustoperatewithin constraintsmposed

by the humanvisualsystem e.g.,contrast,esolution, rightness ,tc.)and he properties f the

10

8/4/2019 97-18


interface e.g., size of knobs,button sizes,and resistance)must mesh with pilot abilities and

anthropomorphic haracteristics.Studiesof biophysical nterfacevariableshave lead to military

standardsor MFDCS design e.g., military standard MIL-STD) -1472D). In particular,a great

deal is known about the responses f the aviator visual systemunder various conditions n

helicoptercockpits (e.g., Frezell, Hohnann, and Oliver, 1973; Frezell et al., 1975; Holly and

Rogers, 1982; Behar, Bachman,and Egenmaier, 1988; Kotulak and Rash, 1992; Rabin, 1995,1996;

Rabin and Wiley, 1996) and about he electro-opticaland physical propertiesof electronicdisplay

devices e.g., Rash, Monroe, and Verona, 1981; Cote, Krueger, and Simmons, 1982; Rash and

Becher, 1982; Rash and Verona, 1989; Kotulak, Morse, and McLean, 1994; Rabin, 1994,1996).

This knowledge is clearly importantbecause t helps ensure hat the various components n modem

aircraftcockpit instrumentpanelshave propertiesconsistentwith crewmembers’biophysical

capabilities.

On the other hand, few research esultsare availablethat define and quantitate he cognitive

dimensionsand problemsrelating o pilots acquiring and maintaining a clear mental picture of the

distributionof information, display objects,menus, data entry fields, and functionsacrosshundreds

of MFDCS display pages; he n-dimensional nterrelationships etween pages;or the most

efficient set of actions o take to navigate o different display pagesor functions. This must become

betterunderstoodso that MFDCS designcriteria can be developed in a truly rational manner.

Surprisingly,developmentof presentand past generationsof MFDCSs have generally been ad

hoc, relying on the experienceand udgment of MFDCS design experts. Most MFDCS designers

organize he information content nto a hierarchicalstructureand then deviate from that structure

when intuition, experience,or testing suggestshat it will be beneficial. The design of an MFDCS

is difficult becauseeven a small contentdatabase an generatean immense number of different

hierarchicalstructures. Searching hrough all the possibilities o find the besthierarchy can be very

difficult, resource ntensive,and time consuming.

Discussionswith membersof currentMFDCS design staffs e.g., at Honeywell, Sikorsky, Army

Research nstitute, U.S. Army Aeromedical ResearchLaboratory) indicate that MFDCS design has

primarily relied on quasi-systematic, onquantitative echniques earned hrough experienceand

validatedwith trial-and-error. For example, Graf and Holley (1988) described he steps aken to

design he MFDCS in the cockpit of the V-22 Osprey. Figure 3 is a schematicof the development

process. The designersstartedwith a mission analysis o determine crewmember duties for the

aircraftand specified range of missions.The designersdetermined how much time crewmembers

had to carry out various tasksduring missions. With this information, the designerscreateda

cockpit design (including MFDCSs). They analyzed he cockpit in two ways. First, they used a

computerizedworkload and performanceanalysis ool to predict whether the current design was

acceptable or the aircraft’s missionprofiles. Second,crewmembers ested he cockpit design n

simulated lights. These man-in-the-loopsimulator flights provided data for determining problem

areas n the design and allowed crewmembers o make commentson positive and negative aspects

of the new cockpit design.The designersmodified the cockpit systemsaccordinglyand iterated he

processuntil cockpit instrumentation apabilitiesmatched mission and usability requirements. As

11

8/4/2019 97-18


canbe magined, hiswasa lengthyprocess.Repeatedlymeasuring ilot-MFDC S interactionn

simulatorss both ime consuming ndexpensive. Moreover,ad hoc changeso the MFDCS (or

otherpartsof the cockp it), hat help solveon eproblem,may inadvertentlyntroduce ew ones.

MSSION MlALYSlS INTERACTIVE ANALYSISp.-.---.-. _~_._._~_._I_~_._ ____._l___.___.-____~-.-.-~-_

II I !-mI

.I

; ;c

i i.

.

k-+

- cOMPurwRED

COCKPIT CRITIW WORKLOAD

DES,GN -, TASK + PERFORM4hlCE

; i.

4tWLEiI.S AMLVSIS

. AWAS)

-_--_-----------

h@N-INTHELOOPSIMULATION

.I

;

i.

I

;

i

OIJTPLITEEDS

l

I-DESIGNRQMTS

-c+i+xmx

;_

I-SAT

;

-M4WRIM

4 1 s‘ Lsq LS AF I

i-ETc

,

I

;

iII

Figure3. The design rocessor the development f the V-22 Ospreycockpitan d

MFDCS. Designersdentify constraintsmposed y mission nalysis nd hen

iterativelybuild a cockpit h at satisfieshose onstraintsmodified rom G raf and

Holley, 1988). The box with the thick edge n dicateswhere he proposed

quantitativemethodwill influence he design roces s.

Published escriptionsf M FDC S design echniques mphasizehat he layout of fun ctions nd

pag es hould ollow general uidelines, ut hey do not explainpracticalmeth ods or satisfyingheguidelinesCalhoun 1978;Lind, 1981;Spiger ndFarrell, 1982 ;MIL-STD-1472D; Williges,

Williges, andFainter,1988;Holley andBusbridge, 995). Someof theseguidelines re:

1. Frequently sed unctions hould e the mostaccessible,

2. Time critical functions houldbe he mostaccessible.

3. Frequently sedand ime critical unctions hould e activated y the buttonsthat eel “ideally located” e.g., op of a columnof buttons).

4. Program epeated election f the samebutton. For exam ple, ocate he most

commonly electedunctionof a menuon the samebutton hat calledup that

me nu. Failing that,program omm on unctionso adjacent u ttons.

12

8/4/2019 97-18


5. The number of levels in the hierarchyshould be as small as possible.

6. The overall time to reach functionsshould be minimized.

7. Functions hat are used ogethershouldbe groupedon the same or adjacent

pages.

8. Related functions on separate agesshould be in a consistent ocation.

9. Related functions should be listed next to each other when on a singlepage.

10. Consider he types of errorscrewmembersmight make and place functions

accordingly o minimize the effect of those errors.

11. In some cases, requently used and time critical functions should be removed

from the hierarchicalstructureand be given dedicateddisplays.

Many of thesegeneralMFDCS design guidelinesare the same as those for structuring he layout

of physical controls Sandersand McCormick, 1987), while others (4,5,8,9, and 11) appear o be

unique to the design of software generated unction selectionswitches or computer-drivendisplay

units. Some of theseguidelines have been investigatedexperimentally. For example, Snowberry,

Parkinson,and Sisson 1983) showed hat searchspeedand accuracy ncreasedas the number of

levels n a hierarchyof user-activated unctionsdecreased 5). Likewise, Teitlebaum and Granda

(1983) demonstrated hat placing related unctions n inconsistentpositionsresulted n a 73 percent

increase n search ime (8). A literaturesearch ound no reportsdocumenting the degreeof

effectivenessof the remaining guidelines,although hey seem reasonableand have face validity.

MFDCS designersselect he guidelines hey consider o be most important. For example, n the

developmentof the MFDCSs for the SuperCobraattackhelicopter, Holly and Busbridge 1995)

focusedon guidelines 1,2,5,7, and 8. The designersgroupedrelated functions into one of eight

subsystemswhich were assigned o the buttonsalong the bottom of the MFDCS as n figure 2).

These were further organized nto two major subgroups.Related information on the samedisplay

page was functionally grouped, and the same nformation on different pageswas presented n thesameposition across he pages. The designersalso emphasizeda minimum-depth approachand

ensured hat all critical information was no more than two levels from the top of the MFDCS page

hierarchy. The most critical information needed o fly and fight was no further than one level from

the top of the hierarchy.

However, applicationof these generalMFDCS designcriteria is problematic because hey often

conflict with each other. For example, shoulda frequently used function be placed by itself near

the top of the hierarchyof the MFDCS pages 1) or should t be placed in a submenuon a

secondarypage with its related, but infrequently used, unctions (7)? Likewise, should criteria3,4

or 7 dominate selectionof a soft-key for a specific unction? Currently, there does not appear o be

a quantitativemethod of deducing the optimal trade-offsso designers ry out different optionsuntil

the entire system“feels” good. This is a time consuming ask becausemovement of a single

function can require a cascadeof related changes hroughout he MFDCS.

With an ad hoc, intuitive, or trial-and-errorapproach o the design of MFDCS data contentand

functionality, operational estsmust be used o judge the performanceof an MFDCS. However, it

13

8/4/2019 97-18


often equires greatdeal of effort both o build new MFDC S layouts nd o m easureheir

performancexperimentally.Designers,herefore,may not havesufficient ime or reso urceso

generate ndvalidatemany alternative es igns. ndeed , n the design f the SuperC obra ’s

MFDCSs,Holley andBusbridge1995 ) conclude, A rap idprototyping apability or control-

display orm ats s suchan important ool that the design f a ‘glasscockpit’shouldnot be

undertaken ithoutone.” Thosedesigners ad accesso simulators nd graphics orkstations, ndso couldquickly ry differentMF DCS configurations. owe ver, here s no indication hat hey

hada quan titative ptimizationmethod or assigningunctionso MFDCS pages ndbuttons. n

figure3, the bold box sugg ests h erea quantitativemeth od or buildinga MFDC S hierarchywould

contributeo the overallcockpitdesign rocess.A quantitativemethodof designmight alsohelp

clarify the relative m portance nd nterrelationshipsf the design uidelinesistedabove.

QuantitativeMFD CS designmethods

Navigating hrough hierarchy f displaypages ontainingunctionsmapped o hardor soft

keys s a commo nly equired ask or many familiar applicationse.g.,automatedellers,computer

programmenus, elephone nsw ering ystems).This section umm arizesomepreviouslydeveloped ene ric ormu las or analysis n d design f hierarchical atastructures. ecause,o

date, hesemethods avenot been ully develope d r validated,hey aregenerallyunsuitableor

complex ractical pplicationsike the designof MFD CSs. This section lso ntroduce s otation

for use n subs equentections.

Most MF DC Ss ncorporate ierarchical tructureshat defineorganization f content nd

navigational aths etween isplaypages r modes.Navigation hrough he hierarchy s

accomplishedia the useof navigational bjects uch smenus,ists,an dsoft or hard -keys. n a

simplebranching ierarchy, achscreen ontainsnformation, isplayobjects e.g.,virtual

instruments,auges, ndwarning ights,symbology, nd ext) andsoftkeys or various unctions.

Activatingsoft-keys n the displayor hard-keybuttons n the MFDCS bezelareused o navigate

through he hierarchy o the desired isplaypages aving he desirednformation nd/or urther

selections. he top of the hierarchy s the one page hat s not a selectionrom any otherpage.

From he top page, he usernavigateshrougha sequencef screenshat s unique or each arget

page. Eachpage n the hierarchy s at a level which ndicates ow many screen she usermustgo

through o reach he page.

Figure2 (hottom)show s art of the hierarchical tructu ren the SuperCockpit FDC S. The top

of the hierarchys a dummypage,as t contains o information xcept hoices o um p to other

pages.Many of the buttons t this op-levelpage andotherpages) renot used,but are ncluded n

the hierarchical tructu reo represent utton ocations.The SYS pagepresents om e nformation(not ndicatedn the hierarchy) ndoptions o ump to otherpages,which are ndicated y links to

pages t the next evel. Thesepageswill present ome nformation nd may) provideoptions or

additional ages t the next evel. Thus, eaching he MAINT page rom the top page equireswo

buttonpushes, ne o accesshe SYS pageand another o accesshe MAINT page.

14

8/4/2019 97-18


Hierarchicaldata structures lso are used n computer scienceapplications or databasesorting.

By arranging he contentsof an ordered databasento hierarchical rees, a computer can more

quickly search he database.A number of algorithmsexist to optimize the layout of a database

(e.g., Knuth, 1973; Lorin, 1975). Unfortunately, thesealgorithmswere developed exclusively to

satisfy equirements or efficiently searching hroughstructureddatabases.These early algorithms,

however, did not include mechanisms or optimizing database tructureswith respect o the

numerousand complex details of human-computer nteractions.The database ayout algorithms for

efficient automatedsearching or simple information do not appear o be generalizable o the more

difficult problem of human searches.Nevertheless, he notation for describinga hierarchy s useful

in both situations.

Consider he hierarchy in figure 4a. It consists f n = 3 page levels, (0,1,2) with m = 3 menu

options representedgraphically as the lines emanating rom nodes)possible rom each page

(represented s the circular nodes). Each page, or node, in the hierarchy s indexed as (i, k) which

indicates he level, 0 I J’< n , of each page and position, 0 I k < m' , in that level (n.b., k=O for the

first page or node at each level). The numbers n the hierarchy schematized n figure 4a suggest

this coding scheme. Note that the total number of pages n this type of hierarchy s: 2 m’ .

j=O

It will be helpful to discusshow this notationcorrespondso movement in the hierarchy. The

“parent” menu, if it exists,of page (i, k) is at position (j - l,Lk / ml) , where 1x1 is the largest

integer ess han x (i.e., round x downward). Likewise, the “children” of page (j, k) , if they exist,

are found at positions (j + 1, Cm) o (i + 1,km + m - 1) . Figures 4b and 4c demonstratehow the

notationcorrespondso the positions n the hierarchicalstructure. This notation only describes he

positionsof pages n the hierarchy, it doesnot require hat a page actually containsa function orjump selection.

Some pages n this hierarchy may contain nformation, virtual instruments,or other display

objects, n addition to mechanisms e.g., menu sectionor soft-keys) o jump to other pagesas

constrained y the interconnections. Other pagesalso contain specific functions hat can be

activated. These functions allow the user o interactwith aircraft systems o perform necessary

tasks.

Suppose here are v functions in a database.Let i=O, 1 ...,v-1 index the functions and letv-1

q(i) = (i, k) indicate the position of the function n the hierarchy. Define Q = m(i) as heset ofi=o

page ndicescontaining functions.

With this notation in hand, we can describea simple model of the human-computer nteraction

and show how to minimize expected unction accessime within a restrictedclassof hierarchies. If

15

8/4/2019 97-18


Level 0

Level 1

Level 2

Level

Level

Level

Level 0

Level 1

Figure4. (a) A hierarchical tructure ith three evelsand hreepossible ptions t each

choice oint. The numbersndicatea codingschemehat dentifies he positionof each

optionat each evel. Eachpositioncanbe identifiedas a coordinate air (j, k) , where he

firstnumberndicateshe evel and he second umber ndicateshe positionwithin the

level. (b) The notation dentifying he positionof the parent o page 2,5). (c) The notation

identifying he positions f the childrenof page 1,2).

16

8/4/2019 97-18


each unction, i, is assigned o a unique page in a hierarchy and has a probability of being needed,

pi , and T+) is the time needed o reachpageq(i), then the expected average) ime that it will take

to navigate o a desiredpage containingany randomly selectedseriesof the functions s:

Accurately estimating q(‘) requiresdetailedknowledge about the interactionbetween computer

and human systems. Lee and MacGregor (1985) proposed he following model. Let c indicate the

time needed or a user to read, mentally interpret,and categorizeone option on a display page. Let

s indicate he time needed o strike a key to selectan option once the userknows which option to

select. Let Y ndicate the time neededby the computer o produce he next display. Let mjk

indicate he number of options at page position (j, k) that the user must categorizebefore making

a choice. Then, assuming hat c, s, and Y are constantacrosspages, he time needed o reachpage

(j,k) is:

where the summation s acrossall the levels that the user must navigate, and the sum identifies how

many hierarchy ocations he user must categorizebetween the top page and page (j, k) .

Lee and MacGregor (1985) considered he situationwhere the user accesses ach function

equally often, pi = 1 / v ; each page has he samenumber of options, m; and the user must go

through a constantnumber of pages,n; to reacha function. Then, assuming hat searching hroughm optionsrequires on average)categorizing m + 1) / 2 options before fmding the desired tem,

the expectedaccess ime boils down to

E(T) = ~+r+~(~+‘ .2 1

Given this analysis,one can determinewhether t is better to have a broad design (with many

optionsper page) or a deep design (with many levels in the hierarchy). With all the functionsat the

bottom level of the hierarchy, t is easy o see hat one needsonly

lnvn=-

lnm

levels in the hierarchy. Substituting he right side of this equation for n above and setting the

derivative of E(T) with respect o m equal to zero produces:

17

8/4/2019 97-18


aEtT)v

[

s+r+c(m+1 )/2 1 c =O--

am m(lnm)* + lnm2 1A bit of algebra h ows hat his means:

m(lnm-1)=1+2(S+r).c

Lee andMacG regor 1985) showed hat f a designermeasureshe termsc, s, andr, then he

expectedesponseime canbe m inimizedby selectinghe num ber f navigationor functionoptions

perMFDCS page,m, that satisfy he ab oveequation.Techniques uch sNewto n’smethod anbe

used o estimatehe value of m. For reasonablealuesof c, s, and , Lee and MacG regor ound hat

m rarelygoes boveeight.

Paapan dRoske-Hofstrand1986 ) considered variationon the Lee and MacG regor nalysis y

hypothesizinghat he man ner n which the navigational r functionoptions re grouped n a

displaycouldaffect he time required o selec t n optionon a menupage. When navigationor

functionoptions n a displaypageare grouped,he effectivenumb er f categorizationsor each

menupagedecreases. his canreduc e he overallselection ecisionime or conversely llow a

largernumb er f optionswhile maintaining he sam edecisionime. For instance,with c = 0.25, s =

0.5, andY= 0.5 (secon ds), ee and Ma cGre gor’s nalysis,hat doesnot incorporate ro uping,

suggestsettingm = 8. On the otherhand, Paapan dRosk e-Hofstran d’snalysis hat ncorpora tes

the mproved fficiencies ue o grouping ptions ivesm = 38 .

Unfortunately,heseanalyticdesign esults reoftenof tangential elevance o m any practicalsituations ecause f current imitations n the designmodels.For example, hysical actors uch

as he sizeof soft-keys ndbezelbuttons swell asdisplaysizeand esolutionypically limits the

maximumnumber f optionselections erpage . Additionally, unctionsearch trategies t each

pagewill likely vary betwe en sers ased ponorganization f the content ndprevious xperience

(Vandierendonck,t al.,1988 ). The line of analysis iscu ssedbovealso estrictstself to very

specific ypesof hierarchies:ones hat useall available ey positions n eachpage compareo

figure2) andwhereall the functions re on the owes t evel. Thus ,evenoptirnality rom Lee and

Mac Greg or’s pproachmay not lead o the best nformationdisplayoverall. Fisher,et al. (1990)

proposed n expanded ch emeor optim izing he searchor sp ecific unctionsn an informationdisplaysystemwith a largerclassof hierarchies.Unfortunately,heir schemes still too limited n

scope or m ostapplications.

Expectedunctionaccessime is no t the only factor hatcanbe minimized. Roske-Hofstrandnd

Paap 1986) described methodof buildinga hierarchical tru cture onsistent ith a u ser’s

“cognitivema p” of the contentdatabase.Subjectsated he similarityof all pairsof the 64 pagesn

18

8/4/2019 97-18


a database.They converted hese similarity ratings nto distancesbetween pages.These values

were then used n an algorithm to solve for a hierarchicalstructureof pageshaving minimum

accessime paths. The resulting structure mproved performance elative to an already existing

hierarchy.

Roske-Hofstrandand Paap (1986) demonstrated he importanceof consideringa user’s mental

model of the relationshipsbetween functions,but it is difficult to design hierarchieswith this

techniquebecausea generally acceptable nd validated measureof a user’s mental model has not

been developed. While a requirement o satisfya similarity relationshipbetween functions seems

to be a useful constraint or designinga hierarchyof displays,other measuresof how functions

complementeach other (e.g., measureof sequentialuse) could also be formulated into valid design

constraintshat would act to offset or exploit relatedcognitive or user nterface imitations or _

advantages.Even if designerscould find a consistentlyaccuratemeasureof cognitive distance

betweenpage contents n a database f information displays, t is not clear how one would build an

appropriatedisplay page hierarchy o minimize that distance. Seidler and Wickens (1992) showed

that cognitive distance nteractedwith other aspectsof a hierarchicalstructurebesidesapparentdifferencesand similarities. Thus, design of a hierarchyof display contentmust take multiple

constraintsnto account. The method used by Roske-Hofstrandand Paap (1986) is too limited in

scope o deal with such additional complexity.

Current state of MFDCS design

The literatureon human-factorsaspectsof MFDCS use and design suggests everal conclusions.

Accessing nformation from MFDCSs with large databases f display page content and user-

selectable unctions can contributesignificantly to crew workload. The design of MFDCS display

page contentsand hierarchiesby industry eaders n avionics seems o be most frequently

performedby applying general“common sense”guidelines hat experienceddesigners mplement

in an ad hoc fashion. A quantitativemethod of balancing he previously listed guidelines for

MFDCS design could help designersdevelop MFDCSs that have higher probabilitiesof having

high function searchefficiency and would have the potential of reducing MFDCS-associated

workloads. Current quantitativedesignmethods for information display systemsseem to be

inadequate.

An investigation nto the designof MFDCS hierarchiesof display pagesor modes and embedded

functionsshould have at least wo principal foci. First, a quantitativemethod of designing a

hierarchyof MFDCS display pagesmust be elaborated hat incorporates s many human factor and

user nterfaceconstraintsand capabilitiesas possible. Without a quantitativedesign tool, designers

of MFDCS page contentsand accesshierarchieswill continue o rely on intuition, luck, inefficient

trial-and-errorexperience,and reports rom the field regardingoperationalproblems with

MFDCSs. Substantialamountsof time and resourcesmay be expendedgeneratingwhat

quantitativemethods might show to be suboptimalhierarchicalstructures hat could be problematic

for pilots in certain high-stresscircumstancese.g., in-flight emergencies). Moreover, without a

quantitativeMFDCS design method, results rom relatedhuman factor studieswill have little

19

8/4/2019 97-18


influencebecausehere s no way to ensurehat the hierarchy eflects he relative mpo rtance f

factors ound o be relevant o effectiveuseof MFDC Ss. The nextsection es cribes quantitative

metho d hat s capable f generating hierarchyof displaypages nduser unctions hat will be

optimalwith respecto designer pecified riteria.

The second lement eeded o advancemodel-based ethodsor organizingMFDCS pagestructuress additional xperimen tal tudy o identify andquantitateherelevantcomponentsf

pilot-MFDCS nteractions. utureMFD CS research lsoshould nvestigateigorously he

previouslyistedMFD CS designguidelineso determinehe extent o which hey adequately

describe ndproperlyweightcognitive actors nd mpo rtant spects f the user nterface. Such

studies ill be requiredo identify realisticvalues andvariances)or the parametersn the

optimization quations. he hum an actor-related aram eterslsomay be parameter&dby user

characteristicse.g.,age ange,gender, xperienceevels,education, r useof performance

enhancingmedications). ikewise,valuesquantitatinghe characteristicsnd performance f the

physical om ponentsf the MF DCS couldbe stratified y specificmanufacturersnddisplay

systems.

A new quantitativemetho d or ontimizingMFD CS contenthierarchies

This sectiondescribes hat we believe o be a new methodof optimizing he hierarchyof

contentpages nd user unctions or MFD CSs .

First, n order o quantify he numerous uman actorconstraintshat could be imp osedduringthe designof the displaypagestructureor an M FD CS , define an overall cost or a givenhierarchyas a weighted inear combinationof an arbitrarynumberof cost unctionsdevelopedosatisfy elatedcriteria:

i=l

Eachconstraint,, imposes cost Ci ) and weights hi) eachcostaccordingo its significance s

obtainedby the designer rom hum an actor expe rts amiliar with the capabilities nd imitations

of the aircraft or which he MFD CS will be installed. The following sectiondescribes ow to

efficiently calculate cost or expected cces sime. Subsequentections emonstrate ow to

selecta hierarchy hatminim izes he cost un ction.

Cost as expected ccessim e

Defining a cost u nction or optimizingan MFDCS pagehierarchydesign equires nowingwhich of the many physicaland software-related roperties f an MFD CS can have significant

effectson performance f re quired n-flight duties. Also, one needs o consider hat some

20

8/4/2019 97-18


MFDCS pages can show either functions or option menus, but not both. Other MFDCSs (as in

figure 2) can simultaneouslydisplay both functions and option menus. For the following

discussion, t is assumed hat the MFDCS is similar to those in figure 2 and portrays functions,

which allow data or control inputs by the user, and menus simultaneously selecting a menu

option typically causesa jump to another display page).

As noted above, a designer may want to minimize the expectedaccess ime acrossall

pages, so C, might be:

V-l

Cl = c&Pii=o

where, as before, the time to reach page (j,k) is:

For a nonhomogeneousprobability distribution, calculating Tik requiresmore effort. To

simplify matters, assume hat userssearch he options on a menu page one at a time, and that the

pages are searched n the order of their indices. Thus, at page (i - 1,Lk nz’1 , a user must

categorizewhichever pagesbetween (i - I + 1,1 / m’] m) and ( - I + 1,1 / m’’ 1) contain a

desiredmenu option. The last page is the option that the user must select o reach page (i, k ) .

(While this is not likely a valid model of how userssearchan MFDCS menu page, the following

analysisdoes not depend on the user’s searchmethod, only that the designer can identify the

method.) It is easy to check for a function at any of these positions by determining if the page inquestion s in the set of function position indices Q. However, if a page is not in Q, its contents

may still need to be scannedand interpretedbecause t could contain a menu choice whose

descendants re function pages. Such a page would have a label that must be categorized. There

is a recursive algorithm that considers hese possibilities. Define the following function:

Ih+m-I

1 if (.L k) E Q or c Htj+l)h > 0h=km

Hjk =

otherwise,

which returnsa value of one if page (j, k) is either a function or is a menu selection that

eventually reachesa function page. The summation simply checks o see if the children of page

(i, k) are function pages or have children that are function pages. Calculation of the H term

21

8/4/2019 97-18


works ts way dow n o the bottomof the hierarchyand hen ilters back up to the top in a

recursive ashion.

The numberof options hat mu stbe mentally categorized t menuposition (*--l,[klm’) is

then:

m(j_/)[k/m’J = Lc ’ H(j-,+l)h ’

h=Lklm’J m

Although he notation s ratherawk ward o look at, it is a relatively simplematter o write a

computer rogram o carry out thesecalculations.

With these orm ulas, t is possibleo calculate he expected cces sime for any layout of

functionson a given hierarchical truc ture. n theory, one couldconsider very possible ayout

of functions nd select he one with the lowestcost. In practice, uchan approachwill rarely

work becausehe numberof possibleayouts ypically will be astronomical. n the followingsectionswe discuss everalnum erical echniquesor solvingcostminimization problems . The

hill-climbing technique s discussedirst and subsequentlyimulated nnealingwhich works

better or cost unctionshavingnumerousocal minima.

Hill-climbing

When d ifferentiableequations,rom w hich analyticaloptimization e sultscanbe directly

obtained, annotbe formulated, om puter cientists ften apply a num erical echnique alled hill-

climbing o find a global max imum or large com plexsystems.After se lecting n initial

MFDCS pagehierarchy,a d esigner an calculate ts cost C(0) using he equations bove.If the

designermodifies he hierarchyand calculates new cost C(1)so hat C(1) c C(O), then the new

hierarchyhasa smallercostand should ep lace he older hierarchy. Iterating his process ill

eventually ead to a hierarchy or se tof hierarchies)or which he costcannotbe reduced ny

further. This approachs calledhill-climbing becauset is analogouso climbing a hill by

moving n whateverdirection s up relative o your currentposition.

An examplewill demonstratehe procedure.Suppose ou want to distribute =5 functionson

the hierarchy ramewo rk n figure4 to minimize C, . Supposehe probabilityof accessing ach

function s:i+ l

Pi==>

so that functionswith higher ndicesare accessed ostoften. To apply the hill-climbing m ethod,

calculate he costof an initial random ayout of the functions.Pick a function at randoman d

randomlypick a page n the hierarchy tructure.Move function to that page and if a different

22

8/4/2019 97-18


function s alreadyat that page,have he functions wappositions).Recalculatehe costand

accept he changef the costdecreases.f the cost nc reases r stays he sam e, evert he system

back o its layoutbefore he move . Continue his proces s ntil the system tops hanging.

Figure5 show s he effect of the hill-climbing procedure.Figure5a shows n initial random

layout of functions.The layout s not optimalandhasa costof C, = 0.587. Figure5b shows he

effect of the first move hat decreasedhe cost. Function3 movedup a level. This reduc es

search ime for that functionwithoutaffectingany other unction’s searchime, thereby educing

cost o C, = 0.513 Figures5c-f show he effectsof subseq uent oves eading o decreasesn

cost. Figure5f show s he final hierarchy esulting rom this procedure.The programstoppedafter one housand onsecutive oves ailed to decreasehe cost. The fmal hierarchyplaces he

mostprobable unctionat the top, the next mostprobable unctions t level 1, and the leastprobable unctionat level 2. This is an optimal ayout or this situation. Figure5g show s he

final hierarchywith non-needed ages em oved.

Cost or related unctions

The designof an MF DC S m ay need o consideractorsother han expected ccessime. Fo r

example,guidelineseven rom page 13 suggestshat the designer houldplace elated unctions

on the samepageor on adjacent ages i.e., if not on the samepage,onebutton-pressway).

The relatednessf two functions andi, R, , canbe estimatedhroughpilot surveys r by

MFD CS design xperts.

Define the page-distance , j, betwee n wo functions, andj, as he max imumnumberof

levelsup one mustgo from either unction o find a menupage hat is parenttoboth functions.

Page-distanceanbe calculatedn the following way. Let q(i) = (1,k) and q(j) = (1+ r,h) with

Y 2 0 sothat function is at the sameor low er level as function . Then the pagedistances:

flj = r + minu E[O,Z]suchthatl-$-]=I--&]}.

As u step s p from 0 to I, the calculation n the right steps p from a child to parentpageand

checks o see f the pathwa ys f the two functions’page s ave converged .The page-distances

the smallest um berof levelsup for which he two pathways onverge.For example,when

yY = 1 either he two functions an be reached rom the sameparentpageor one functioncanbe

reached y a selection rom the otherpage. Minimization of the following cost erm will putrelated unctions s closeas possible:

v-l v-lC, = ~,~,RuH$ .

i=O=O

23

8/4/2019 97-18


Figure 5. The development of a hierarchy through hill-climbing. (a) The initial layout of

functions produces a high cost. (b) Function 3 has moved up a level. This reduces he

number of stepsneeded to reach the function. (c) Function 0 has moved to the left. This

frees a menu label at level 1, and reducescategorization time on the way to functions 1

and 2. (d) Function 2 moves up a level. This reduces the number of steps needed to

reach the function. (e) Function 1 moves up a level. This reduces he number of steps

needed to reach the function. (f) Functions 1 and 2 swap positions. This places the more

probable function in a position to be categorized first. Further changes do not reducecost. (g) The fmal hierarchy with non-needed pages removed. [The following parameters

were use: c=O.1,1-o. 1, s=O.2.]

24

8/4/2019 97-18


To dem onstrateow this cost erm work s, et relatedness etween unctions bey he

following formula:

1

1 if Ji-jl <2

R, =

0 otherwise.

This me ans hat functions and 4 arerelated,2 and 3 are related,but functions and4 and 1 and

3 are not related. Figure6 showshow the hill-climbing procedure tartswith an initial layou tof

functions a) a ndchangeso a hierarchy hat places elated unctionswithin one buttonpress f

eachother. For this particular xample, he system eededonly two moves o reachan optimal

layout. In (b) function1 movedup a level, therebyplacing t closer o function0, while keeping

it the sam edistanceo function2. In (c) function4 moved o the right, therebymoving t closer

to function3. Th is optimal ayout s not unique; igure 6d show s very different ayout that also

minimizesC, .

Figure6. The developm ent f a h ierarchy hat minimizesdistance etween

related unctions . a) The initial layout of functionsproduces high cost. (b)

Function1 movescloser o function0. (c) Function4 movescloser o function3.

(d) Another ayoutof functions hat has he sameoptimalcost.

25

8/4/2019 97-18


Cost or expected ccessime and relatedness

Generally,a designerwill want to build an MFD CS that satisfiesmany different constraints.

For example, f the MFD CS hierarchyshouldminimize both expected ccessime andput

related unctions lose og ether, he cost o be minimizedwould be:

c= c, +c,.

Figure7a showsa hierarchyproduced y the hill-climbing method or this cost. The hierarchy

doesa good ob of minimizing both cost erm s. Every function s within one pageof its related

functions, nd he mostprobab le unctions re ocatedat the highest evels. How ever, h e layout

in figure 7a is not optimal. Figure 7b shows n o$rnal layout, foundby the hill-climbing

method with different starting conditions. Here, each function is within one page of related

functions, and function 1 is placed at level 1, thereby reducing the expected access ime.

Figure7. Hierarchies or minimizationof expected ccessime and relatedness.

(a) A layout of functions often found with the hill-climbing method. This layout

cannot be modified to produce a lower cost.(b) An optimal layout of functions

found occasionally with the hill-climbing method.

26

8/4/2019 97-18


Note too that the optimal ayout s not the sam e s he optimal ayout or expected ccessime.

To m inimizeexpected ccessime, functions1 and2 need o switchpositions nd unction0

needs o move under unction3. The latter s not possible ecaus eunctions and 1 are related

while functions and 3 are not. Thus,placing unction0 beneath unction3 would ncrease ost.

With that constraint, xpec ted ccessime is shorterwhen function1 (and ts child 0) is reached

through he second ositionof level 1 rather han hrough he third position as n figure 5). Thisensureshat a userspendsess ime searchinghroughoptionswhile accessinghese wo

functions.Thus, mposingmultipleconstraintsequires tructuringhe hierarchyquite

differently han might be expectedrom imposing itherconstraint y itself.

It is important o realize hat he hill-climbing methodcannotmodify the layout of functions n

figure7a to produce n optimal ayout. The move ment f any functionwill lead o an increasen

costandwill be rejectedby the hill-climbing technique .This layout,or state,of the system scalleda stablestate. No singlemove will cause he system o modify itself. A non-optimal tab le

state,where herecanbe no furtherdecreasesn c ost, s a local minimumof the cost. The problem

with a hill-climbingmethod s that t c anneveraccept change hatmight ncreasehe overallcost.As the above xampledemonstrates,ometimeshe systemmu st oleratencreasesn cost o reach

oneof the globalminima. Researchersavedevised numberof methodsor resolving he

problem, nd he next section escribesne of the mostgeneralmethods.

Simulated nnealing

Simu lated nnealings a techniquehat allowshill-climbingprocedureso avoid he local

minimaof a cost unctionandsearch ut a globalminimum. Intuitively,eachpossibleayoutof the

functions, r state, orrespon dso a position longan axis ine (actually t is a position n a higher-

dimensional pace).The cost unction t eachposition long he axisdefines curvealong he line.

Changing tateslayoutof functions)s then ike movingalong he ine. H ill-climbing echniquesstart t an initial point on the ine andmove n a direction hat goesdown he costcurve figure8a).

As a resu lt, he technique anbecomerappedn localvalleys.

27

8/4/2019 97-18


(4 d

Figure8. An intuitive description f hill-climbingand simulated nnealing. he

layoutof functionss like a ball on a hill (c ost). Movementof the ball dow n he hill

correspondso changesn the ayout hat ead o lower costs. a) A hill-climbing

method anbecome rappedn localminim a. (b) With simulated nnealing t a

high emperature,he ball often ravelsuphill andout of local (and global)minima.

(c) With sim ulated nn ealing t an intermed iateemperature,he ball can ravelout

of localminima but not out of the deeper lobalminima.

Simu lated nnealing voids hisproblemby introducing xtra“energy” nto the optim ization

process.With this meth od, nitially, movem ent long he costcurvecanoccur n directionshat go

up or down. Then the prob abilityof movingup he costcurve s gradually educed s ime (or

iterations) rogresse s. s the likelihoodof movingup the costcurvedecreases,he systems more

likely to becom e tuck n the deepe st alley of the curve,a globalminimum. It is mo re ikely to

climb out of localminima becausehey arenot asdeep.

Formally,definea temperature, , w hich starts t a largevaluean dgraduallydecreases. uppose

a random hangen the hierarchy t time t produces cost, C(t). Accept he changewith

probability:

28

8/4/2019 97-18


if C(t) < C(f - 1)

P=

Ixp( C(t) / T)

1+ exp( C(t) / T) Otherwise*

Changeshatdecrease ostarealwaysaccepted,ndchangeshat ncrease ostareaccepted ith a

probabilityhat depends n the costof the new ayoutand he currentemperature.When T is

large, he exponentialermsareclose o one,and he probability f acceptinghe changes close o

onehalf, even f sucha changeeads o an ncreasen cost. As figure8b schematizes,his allowsthe systemo m ove out of localminima,bu talsoallows he systemo move out of globalminima.

As T decreasesn size,a largecost en ds o make he probabilityof acceptinghe change lose o

zero. By gradually ecreasing , the system oes hrough phasewhere t becom es tuck n a

valley of the costcurve hat containshegloballyminimumcost,but s able o climb out of valleys

in the costcurve hat containhighercostsfigure8~). As T decreasesurther, t rem ainsn thegloballyminimumcostvalley.

Simu lated nnealing equires substantialmountof computation.The technique equires

startingwith a large nitial tempe raturend slowly decreasingt, all while making changeso the

hierarchical tructure.Selecting he initial temperature nd he rateof decreas es important. f

the temperatures too small nitially or decrea sesoo quickly, the systemwill become tuck n a

local minimum. On the otherhand, f the temperatures very largeanddecreasesery slowly,

the systemwill spendmuchof its time acceptingandomchanges ndwill take a very long time

to produce final solution. While bounds xiston both the startingemperature nd on the rate

of annealing,hey tend to be impractical or use Gem anand Geman,1984). For the simulations

reported ere, he initial temperatu re a s T(0) = 3900, and t then decreas ed ith every random

moveof a functionas:

T(O)T(f) 10+t *

As w ith the hill-climbing procedure ,he algorithm erm inatedwhenone housand onsecutive

moves ailed to produce decreasen cost.

Figure9 compareshe costof solutionsoundby the hill-climbing methodwith those ound

usingsimulated nnealing. One hundredrialswererun for eachapproach.Figure9a shows he

frequency f differentcosts oundwith the hill-climbing method. Most of the solutions ave a

costof C=8.380, althoughoccasionallyhe system inds the optimalsolutionwith C=8.353.

Figure9b sh ows he analogousesults or simulated nnealing. t finds an optimal ayout much

moreoften han any other ayoutbut doesoccasionally onvergeo non-optimalhierarchies.

29

8/4/2019 97-18


SO

8.35 8.4 8.45 8.5 8.55 8.6 8.65 6.7 8.75 8.8

C;OSt

83

ho!iE40IL

20

08.35 8.4 8.45 8.5 8.55 8.6 8.65 8.7 8.75 8.8

cost

Figure9. Frequen cy f final hierarchy osts or hill-climbing a) and simulated

annealing b).

30

8/4/2019 97-18


Simulated nnealings not the only option or costminimization. Other echniques xist hat

may do ust aswell or be tter. The main observation ere s that a cost unctioncan measurehe

qualityof a hierarchical tructure.Minimization of that cost unctionallows a designero

identify an optim alhierarchy. The most mportant art of this proces ss identifying how to

convert he qualitativeguidelines or hierarchical esign nto costs.The next section onsiders

this ssue n m ore detail.

Cost unctions

The previous ection escribed generalmethod hat finds a hierarchical tructureo minimize

a cost. This section onsiders ow to quantify he qualitativedesignguidelines o produce osts.

Frequently se d unctions

Guidelineone suggestshat frequentlyused unctions houldbe placed n the mostaccessible

locations.This is alreadyaccom plished y the equation or C, above:

v- l

CI = C Tq(i) Pi -i=O

As figure5 demo nstrates,inimizationof this cost erm ends o push he functions sedwith high

probability o the top of the hierarchy.Probability stimates anbe gathered ither hrough xpert

opinion,pilot interviews, r datacollection uring lights.

Time critical functions

Guideline wo su ggestshat time critical functions houldbe placed n the mostaccessible

locations.Minimizing the following equationwill apply his guideline:

v- l

C3 = c Tq(i) i .i=O

Thisequations the same s or C, with importance,i replacing robability. Mimmizationof

thiscost ermwill p lace he functionswith high mportance t the top of the hierarchy. Expe rt

opinionor pilot interviews anprovideestimates f function mportance.

Ideal ocations

The third guidelinesuggestshat frequentlyusedand ime critical functions houldbe

activated y buttons hat feel ideally located.Assum e hat the button ndices,h = 0,. . m - 1

correspondo the relativeorderof pageselection uttons n eachmenupage n the hierarchical

31

8/4/2019 97-18


framework,and that the lower the index of the button, he higher ts “idealness.” Pag e (j, K) in

the hierarchywill then usebutton:

b(j,k)= k-Lklmj m.

Let the time increase ue o beingnonidealbe some ncreasingunction j[h] . Then, duringnavigation hrough he hierarchy o reachpage (j, k) , the sumof time increase s t eachbutton

push s:

Bjk = gf[b(j-I,lk/dJ)] .

I=0

This summ ation oesbackwardshrough he hierarchy rom page (j, k) to the top and dentifies

the buttons ecessaryo reachpage (j, k) . This time shouldbe added o the overall accessim e

needed o reachpage (j, k) . Thus, he cost erms or C, and C, shoulduse he following

equation or accessime:

Quantificationof the term “idealness” s neede d efore he effect of o rderon buttonselection an

be modeled. Presumably,deally ordere d i.e., allows user o m ostefficiently reac hneeded

functions)pagebuttons nd their identifierswill be categorized ore quickly, search edaster,orstruckmorequickly. Experimental tudies houldbe able o determ ine he influenceof ideal

buttonorderingand delineate he approp riate efinition or value able for f[h] . Then the

algorithmelucidated bovewill allow designe rso optimizeactualMFD CS displaypage

hierarchieso take advantag e f theseadditionalhuman actorsMFDCS-userperformance ata.

Reneated electionof buttons

Guideline our suggestshat the hierarchystructure houldminimize the need o switch

buttons or the most requentlyused unctions. Thus, he mostcommonselection rom a m enu

shouldbe on the samebuttonascalledup that menu. Presumably w itchingbuttons dds o the

overall time for the user o respond ecause e m ustmove his finger to a n ew location. Let the

time to travel a unit distance e a. Then, while navigating rom the top of the hierarchy o p age

(j, k) , the time sp entmoving betweenbuttonswill be:

sjk =afJd[,(j-Z-l,Lk/m”*“J ,b(j-Z,~kim’~)] .I=0

Here , he firstb term s the buttonassociated ith a higher evel and he second term s the button

associated ith the subse quentevel on the way to page (j, k) . The function d[ ] is a measure f

32

8/4/2019 97-18


the physical distancebetween the buttons. The summation measures his distance acrossall the

selections hat the user must make to reach the desiredpage.

The time spent moving between buttons should be added to the calculation of access ime for

the costs C, and C, . The access ime to reach page (i, k) should now be:

where s, which previously measured he entire strike time, now incorporateswhatever partsof the

strikemovement are common to all keys.

Minimize number of levels

Guideline five suggested hat a hierarchy should have as few levels as possible. Too many

levels could lead to fatigue or cause he user to become lost. A simple measure of depth would

be to add up the level indices of all function positions:

v- l

c4= cJIq(i)]i=O

Here,J[q(i)]e ers to the level index at the position of function i, q(i) = (j,k) . When many

functions are at deep levels, C, will be large. It appears hat current MFDCS designs consider

this cost to be very important (Holley and Busbridge, 1995). One easy method of minimizing the

number of levels in the MFDCS hierarchy is to provide a large number of buttons. However,

with such an approach, he user trades he searchof the hierarchy for the searchof the proper

button. A cost for such a searchcould easily be included in the method describedhere.

Minimize overall access ime

Guideline six suggests hat the hierarchy should minimize the overall access ime. Minimizing

C, already applies this guideline.

Related functions on close pw

Guideline seven suggests hat related functions should be placed on the same page or on

adjacentpages. Minimization of Cz places functions as close as possible.

Consistent ocation of related items

Guideline eight suggests hat related items should be in a consistent ocation, acrossdifferent

pages. Assuming that being close to each other corresponds o being close to each other among

33

8/4/2019 97-18


the buttons,minimizationof the following cost unctionwill put related unc tions s closeas

possible:

With the relatednessunction,Rti, as defined n the section Cost or related unctions.”

Related unctions n the sameDa=

Guidelinenine suggestshat, when hey are on the samepage, elated unctionsshouldbe

placednext o eachother. Minimizing C, alreadyapplies his guideline.

Errors

Guideline en sugg estshat the hierarchyshouldanticipate ikely errorsand minimize the

effect of thoseerrors. A full enume ration f likely errorsand he manner n which they dependon variousaviation-related uman actorswill requireextensive xperimentalwork. Such

researchmustconsider t least: the physical ayout of b uttons, he labels or o ptions, he effects

of fatigue,and effectsof aircraftvibration. f design ers an quantitatehe relationshipof various

variablesor factors hat predictdifferent ypesof errorsassociated ith MFDC S use, hen a cost

functioncanbe determined uch hat its value will be large whenan MF DC S function,or group

of functions,s in a high errorrisk location n the hierarchy.A precisedefinition of this costterm will depe nd n the analysisof errorsand heir relative operationalmpacts.

Dedicateddisnlavs

Guideline 11 sug gestshat some requentlyusedand ime critical functionsshouldbe removedfrom the MF DC S and given dedicated isplays.Consideration f this issuedoesnot requirea

new cost unction. The cost unctionsdiscussed reviouslywill optimize nformationacross

multiple MF DC S hierarchies imultaneously ndplace unctionsn separateMFDCSs. The

designer eedonly specify h e num berof M FDCS s, the numberof levels for each ,and he

numberof menuoptions or each. A d edicated isplaywould simplybe an MFD CS with one

level and one option. With the cost unctionsdescribed bove, educing he overall costshould

place he mostcommonlyusedand ime critical functions n the dedicated isplays,secondary

information n the MFD CSs , and placerelated unctions n the hierarchyof a comm onMTDCS.This approach lsocould decidewhich functions o placeon a HUD.

34

8/4/2019 97-18


Discussion

This paper describeda new quantitative method for optimally distributing control and input

functions acrossa hierarchy of MFDCS display pages. The flexible method is based either on

minimizing a single composite cost function that is a weighted linear combination of separate

cost functions or minimizing a set of simultaneouscost functions. Such cost functions can

accommodatean arbitrary number of design or pilot performance constraints. We illustrated

how these cost functions can be developed from qualitative MFDCS design and human factors

constraints. Cost function coefficients are the elements in the equations hat qua&ate the

effects of operationally important MFDCS human factors. It is important to emphasize that the

method describedhere does not necessarilyguarantee deal hierarchical MFDCS display

structure or all conceivable situations. A designermust still verify that the quantitatively

determinedhierarchy is a good design by experimentally demonstratingbetter performance than

nonoptimized baseline designs n realistic scenarios. The proposedcost function minimization

method does find the best MFDCS page hierarchy for the selecteddesign and human factor

constraints. If the cost functions do not adequatelyrepresentor emphasize the factors important

for effective use of a particular MFDCS, the method may not produce a content page hierarchy

that maximizes actual performance.

We anticipate that this design method, when validated, will be a useful design tool that can be

usedto produce optimal or near optimal relationshipsand interpagenavigational paths for

MFDCSs. However, specific applicationswill still need to be augmented with verification

studiesand evaluated in the light of experienceand good udgment since it is unlikely that any

single MFDCS information content design tool could take into account all potentially important

factors or all conceivable operational circumstances. Quantifying the layout of MFDCS

allocated unctions, however, will assistdesigners o more rapidly evaluate the relative

effectivenessof alternative MFDCS display page hierarchiesand better select from alternativehardware-softwarecombinations. For example, the design method described n this report could

be used to obtain an optimal design for an MFDCS that includes both push-button and speech-

recognition as alternative interfaces for accessing dentical information. Since the human factors

and performance parameters or these two disparate ypes of interface would differ, minimization

of an overall cost function or weighted.sum of separatecost functions value could serve as an

objective measure for selecting the best interface. The designerwould optimize the MFDCS

information and control function layout for each interface alone and in combination and compare

total systemperformance for each alternative (Reising and Curry, 1987). Such optimization

would be difficult to perform without the quantitative echniquesdescribedhere.

This report describedsome relatively simple models of the interactionof pilot factorswith the

structureof MFDCS display contentsand distributionof control functions. A designercould

introducemore involved models with no change n the fundamentalcomputational echnique

(althoughsubstantiallymore bookkeepingwould be required). For example, the models discussed

in previoussectionsassumed hat MFDCS computer esponse ime was constant or all functions.

That is probably not realistic, but inclusion of more realism would only require estimating the

35

8/4/2019 97-18


responseime for each unctionand using hatestimaten the appropriate alculationsor access

time. Likewise,usingan interfaceother hanpush-buttons ould equiremodifyingsomeof the

cost unctions nddeleting ndcreatingothers.

Modelssuch s he onedevelopedn this echnical eportcanenhance nderstandingf the

interrelationshipsetween uman actors, he organization f MFD CS displaycontent, nddistribution f interface bjects e.g.,push-buttons)y explicitlydelineatinghe hypothesized

relationshipssequationshat can be solvedwith eitheranalyticor num erical echniques.

Parameter r variablesensitivity nalysis anbe performedo determinewhich parameters rvariables,whenperturbed, ause he greatest hangesn the cost u nction(s).This canassist

designerso focuson the more ntluentialparameters. ikewise,changesn parameters anbe

linked o operational ettings o hatdesignersan dentify the scenariosor which a se lected

MFD CS content ierarchymay be subop timal r problematic.

A quantitative uman actors rientedMFD CS designmodelalsocan assistn identifying

specific nowledg e aps n this topic area hatneedadditional esearch .Suchmodel-directed

research an esult n focused racticalgoal-orientedesearch fforts hat generateesults hathaveimmediate pplication.The modeldevelopedn this reportalsomay play a role in m aking he

MFD CS design ffortmoreefficientand cost-effective y reducing equirem entsor prototyping

an drepetitive xp ensive nd im e-consum ing esign-test-mo dify-retestycles.

Conclusions

Having elaboratedhe generalstructureor a quantitativemethod or incorporating uman

factors nto MFD CS design, ollow-on experimentalwork will be required o establish ealistic

and usefulparameter alues e.g., meansand standa rd rrors) or the coefficients n the cost

functions. Sensitivityanalysismay alsobe performed o qu antify h e relative effectiveness f the

descriptiveMFDC S designguidelines urrentlyusedby experienced esigners.Further

investigationnto the issues resentedn this reportmay alsoresult n the delineationof

additional mpo rtantphysical,cognitive,an dpsychom otor um an ac tors or the efficient and

effectiveuseof M FDC Ss during nflight emergencies r otherhigh workloador high stress

situations.Potential heoretical xpansion f the concep ts numeratedn this report,as well as

the resultsof supporting xperimentalwork, can ead to the eventualdevelopment f useful

quantitative uman actors-oriented FD CS softwaredesign ools. S uchdesignaidsmay lead

to improvementsn aircraftMFDC Ss which allow crewmem berso more efficiently utilize thecapabilities f comp lexaircraft,particularly n em ergency r high workloadsituationswhere

navigating o the required nformationand functionbuttonsmustbe performed apidly andwithout error.

8/4/2019 97-18


References

Behar, ., Bachm an,W., andEgenmaier,W. 1988. C ontrast ensitivitv n Army aviator

candidates: vclonlegiaeffectsandpopulationnorms. Fort Rucker,AL: U.S. Army

AeromedicalResearch aboratory.USAARL ReportNo. 88-16.

Calhoun,G. 1978. Control ogic design riteria or multifunction witchingdevices. n:

o r ceedin f 1Meetin . 383-387.

Cote,D., Krueger,G., and Simm ons,R. 1982. Com parison f heliconter onilot workload

while using hreenavigationsvstems uringnan-of-the-earthlight. Fort Ruck er,AL:U.S. Army AeromedicalResearch aboratory. USA ARL ReportNo. 82-9.

Department f Defense. 19 81. Military standard: umanenpineerine esigncriteria or militarv

system s. quipm ent. nd facilities. MIL-STD-1472D .

Dohme,J. 1995. The military quest or flight trainingeffectiveness.n: Vertical Flight

Training. W. Lars en,R. Randle,andL. P opish eds.)NASA ReferencePublication1373.

Fisher,D., Yungkurth,E., and Moss, S. 1990. Optimal menuhierarchydesign: Syntaxand

semantics. umanFactors, 32: 665-683.

Frezell,T., Hofmann,M., and Oliver, R. 1973. A viator visualperformancen the UH-1H:

Study . Fort Rucker,AL: U.S. Army AeromedicalResearch aboratory. USA ARLReportNo. 74-7.

Frezell,T., Hofmann,M., Snow,A. andMcN utt, P. 1975. Aviator visualperformancen theUH-1: Study I. Fort Rucker,AL: U.S. Army AeromedicalResearch aboratory.

USA4RL ReportNo. 75-l 1.

Geman,S., and Gem an,D. 1984. Stochasticelaxation,Gibbsdistributions, nd he Bayesian

restoration f imag es. EEE Transactions n PatternAnalvsisand Machine ntellieence.

PAMI-6: 721-741.

Graf, V., andHolley, C. 1988. H uman actors mpacton the V-22 Ospreycockpitdevelopment:

An overview. In: Proceed ings f the Am ericanHeliconterSocietv44th Annual Forum.

201-212.

Hannen,M., and Cloud,T. 1995. A case tudy n the design nd es tingof hands-on ontrols:

The LongbowApachegrip development rocess . n: Proceed inpsf the American

HeliconterSocietv51stAnnual Forum. 1417-1435.

37

8/4/2019 97-18


Holley, C., and Busbridge,M. 1995 . Evolution of the venom variantof the AH -l W

Supercockpit. n: Proceedings f the AmericanHeliconterSocietv5 1st Annual Forum.

1436-1449.

Holly, F., and Rogers,V. 1982 . Direct and neiPhboring ensitivitychanges roduced y red

an dblue-whiteadanting ields. Fort Rucker,AL: U.S. Army AeromedicalResearchLaboratory. USA ARL ReportNo. 82- 1.

Knutb, D. 1973. The Art of Com puterPromarnmine,Volume 3: Sortingand Searching.

Reading,MA: Addison-Wesley ublishingCo.

Kotulak, J., and R ash,C. 1992 . V isual acuity with seco nd nd third Peneration ight vision

weles obtained rom a new methodof nieht sky simulationacross wide rangeof

tarpetcontrast. Fort R ucker,AL: U.S. Army AeromedicalResearch aboratory.

USAARL ReportNo. 92-9.

Kotulak, J., Morse, S., and McLean, W. 1994. Does dismayphosphor andwidth ffect he

ability of the eye to focus? Fort Rucker ,AL: U.S. Army AeromedicalResearch

Laboratory. USA ARL ReportNo. 94-34.

Lee, E., andMacGregor,J. 1985. Minimizing usersearch ime in menuretrievalsystems.

HumanFactors. 27: 157-l 62.

Lind, J. 1981. Evaluationof cockpitprocedures. isplays. nd controls or storesmanagement

in the advan ced ircraftarmam ent ystem AAA S). China Lake, CA: Naval Weapon sCenter. NW C TechnicalMemorandum 538.

Lorin, H. 1975. Sortingand SortSystems.Reading,MA: Addison-Wesley ublishingCo.

McBryan, B., and Hall, J. 1995. E ngineering pproachor rotorcraftpilot’s asso ciate ognitive

decisionaiding systems evelopment. n: IEEE/AL% 4 Digital Avionics Systems

Conference roceedings,EEE: Piscataway, J . 76-8 1.

Newman,R. 1995. Head-UpDisplays:Desirming:he way ahead. Brookfield,VT: AshgatePublishingCo.

Paap,K., and Roske-Hofstrand, . 1986. Th e optimalnumberof menuoptionsper panel.

HumanFactors. 28: 377-385.

Rabin, J., and Wiley, R. 1996. Differences n anparent ontrastn vellow and white light. Fort

Rucker,AL: U.S. Army AeromedicalResearch aboratory. USA ARL ReportNo. 96-

21.

38

8/4/2019 97-18


Rabin, J. 1994. Forward-looking infrared: Capabilities for searchand rescue. Fort Rucker, AL:

U.S. Army Aeromedical ResearchLaboratory. USAARL Report No. 94-l 1.

Rabin, J. 1995. Luminance effects on visual acuitv and small letter contrastsensitivitv. Fort

Rucker, AL: U.S. Army Aeromedical ResearchLaboratory. USAARL Report No. 95-14.

Rabin, J. 1996. Image contrastand visual acuity through night vision goggles. Fort Rucker,

AL: U.S. Army Aeromedical ResearchLaboratory. USAARL Report No. 96-26.

Rash, C., and Becher, J. 1982. Analvsis of image smear in CRT disnlavsdue to scanrate and

phosphorpersistence. Fort Rucker, AL: U.S. Army Aeromedical ResearchLaboratory.

USAARL Report No. 83-5.

Rash, C., and Verona, R. 1989. Cockpit lighting comnatibilitv with image intensification night

imagine svstems: Issuesand answers. Fort Rucker, AL: U.S. Army Aeromedical

ResearchLaboratory. USAARL Report No. 89-6.

Rash, C., Monroe, D., and Verona, R. 1981. Computer model for the evaluation of svmbology

contrast n the intemated helmet and disnlav siphtinp svstem. Fort Rucker, AL:

U.S. Army Aeromedical ResearchLaboratory. USAARL Report No. 8 l-6.

Reising, J., and Curry, D. 1987. A comparisonof voice and multifunction controls: Logic

design s the key. Ergonomics. 30: 1063-1077.

Rogers, W., Cabrera, E., Walker, N., Gilbert, K., and Fisk, A. 1996. A survey of automatic

teller machine usage across he adult life span. Human Factors. 38: 156-l 66.

Roske-Hofstrand,R., and Paap, K. 1986. Cognitive networks as a guide to menu organization:

An application in the automated cockpit. Ergonomics. 29: 1301-l 3 11.

Sanders,M., and McCormick, E. 1987. Human Factors n Emjneerinn and Design. New York,

NY: McGraw-Hill Book Co.

Sarter, N., and Woods, D. 1995. How in the world did we ever get into that mode? Mode error

and awareness n supervisorycontrol. Human Factors. 37: 5-19.

Seidler, K., and Wickens, C. 1992. Distance and organization in multifunction displays. Human

Factors. 34: 555-569.

Sirevaag,E., Kramer, A., Wickens, C., Reisweber, M., Strayer, D., and Grenell, J. 1993.

Assessmentof pilot performance and mental workload in rotary wing aircraft.

ErPonomics. 36: 1121-l 140.

39

8/4/2019 97-18


.

Snowberry,K., Parkinson,S.R., and Sisson,N. 1983. Com puter isplaymenus. Ergonomics.26(7): 699-712 .

Spiger,R., and Farrell, R. 1982 . Surveyof multi-functiondisplayand control echnology.

Seattle,WA: NASA-CR-l 67510 .

Stokes,A., and Wickens,C. 1988 . Aviation displays. n: Hum anFactors n Aviation. E.

Wiener and D. Nagel (eds.).SanDiego, CA: AcademicPress.

Teitlebaum,R., and Granda,R. 1983. The effectsof positionalconstancy n searchingmenusfor information. In: Proceedings New York,

NY: Association or Compu tingMachinery.

Vandierendonck, ., Van Ho e, R., andDe S oete,G. 1988. M enu search s a functionof menu

organization, ategorization, nd experience.Acta Psvchologica.69: 23 -248.

Williges, R., Williges, B., andFainter,R. 1988 . S oftware nterfaces or aviation system s.n:

Hum an Factors n Aviation. E. W iener and D. Nagel (eds.).San Diego, CA: Academic

Press.

97-18

Documents