USF Libraries
USF Digital Collections

Operational evaluation of advanced safety enhancement devices

MISSING IMAGE

Material Information

Title:
Operational evaluation of advanced safety enhancement devices rearview video system
Physical Description:
Book
Language:
English
Creator:
Kourtellis, Achilleas
Publisher:
University of South Florida
Place of Publication:
Tampa, Fla
Publication Date:

Subjects

Subjects / Keywords:
Camera
Methodology
Implementation
Experimental
In-vehicle
Driver behavior
Dissertations, Academic -- Civil Engineering -- Doctoral -- USF   ( lcsh )
Genre:
non-fiction   ( marcgt )

Notes

Summary:
ABSTRACT: Since the creation of the automobile, there has been an effort to create and implement mechanical and electronic devices that would improve vehicle safety. In recent years, electronic technologies have become more efficient and cost effective, therefore creating a great spike in widespread implementation. These safety related devices have to be tested for their reliability and amount of help they provide the driver with. The end user (the driver) has to be involved for a successful device. This research presents the methodology used to evaluate the effectiveness of the rearview video system (RVS) used in vehicles, especially in large commercial trucks and effectively the methodology for a more complete investigation of the problem of correctly implementing a safety device. The focus of this research is backing crashes that involve large trucks. The countermeasure tested was a rearview video system which provides a rear view to the driver in real time.A traditional crash data analysis is almost impossible since there is not enough data to perform it, and no data are available for the use of this system since it is fairly new to the market. A driver experiment under controlled conditions was used to create and collect the data necessary for the analysis. The experiment yielded a total of 71 crashes out of 270 maneuvers (26.3%). When analyzed, three backing neuvers yielded different probabilities of having a backing crash with and without the RVS. The increase in stop rate ranged from 46.67 percent to 4.44 percent. This is interpreted as crash reduction due to the device. Driver behavior was observed during the experiment and measured for significant differences. The drivers needed on average 6.47 seconds more time for the maneuvers with the RVS in use. They spent less time looking at mirrors and did it less frequently in order to accommodate the additional glance location presented to them.Overall they seemed to be able to manage their time with some exceptions. The driver acceptance of the device was also measured with a survey given to them after they completed the test. Overall in all measures the majority of drivers agreed that the system helps in reducing the rear blind spot and thus it is a helpful device in reducing backing crashes since it will help them avoid potential hazards while backing. The majority also stated that they would like to have the device in their truck for every day operations. These results show an acceptance of the device and therefore the maximization of the device's use and potential benefits. The RVS is therefore effective in reducing potential backing crashes. The results presented here are limited, and inferences are made with the experiment conditions in mind. General application of the results is possible, with certain assumptions and restrictions.
Thesis:
Dissertation (Ph.D.)--University of South Florida, 2009.
Bibliography:
Includes bibliographical references.
System Details:
Mode of access: World Wide Web.
System Details:
System requirements: World Wide Web browser and PDF reader.
Statement of Responsibility:
by Achilleas Kourtellis.
General Note:
Title from PDF of title page.
General Note:
Document formatted into pages; contains 118 pages.
General Note:
Includes vita.

Record Information

Source Institution:
University of South Florida Library
Holding Location:
University of South Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 002069523
oclc - 608554857
usfldc doi - E14-SFE0003252
usfldc handle - e14.3252
System ID:
SFS0027568:00001


This item is only available as the following downloads:


Full Text

PAGE 1

OperationalEvaluationofAdvancedSafetyEnhancementDevices:RearviewVideoSystem by AchilleasKourtellis Adissertationsubmittedinpartialfulllment oftherequirementsforthedegreeof DoctorofPhilosophy DepartmentofCivilandEnvironmentalEngineering CollegeofEngineering UniversityofSouthFlorida Co-MajorProfessor:JianJohnLu,Ph.D. Co-MajorProfessor:Pei-SungLin,Ph.D. ChanyoungLee,Ph.D. AbdulPinjari,Ph.D. GeorgeYanev,Ph.D. YuZhang,Ph.D. DateofApproval: October14,2009 Keywords:camera,methodology,implementation,experimental,in-vehicle,driverbehavior c 2009,AchilleasKourtellis

PAGE 2

Tomyparents,PanayiotaandIoannisKourtellisandmybelovedwife,Hayley.

PAGE 3

Acknowledgments IwouldliketothankDr.JianJohnLu,andDr.Pei-SungLinforbeingmyco-majorprofessors andacademicsupervisors,fortheirguidanceandadviceduringthepastfewyears.Iwouldalso liketothankDr.ChanyoungLee,Dr.AbdulPinjari,Dr.GeorgeYanevandDr.YuZhangfor servingasmygraduateadvisorycommitteeandDr.EdwardMierzejewskiforservingasthedefensechair.IwouldespeciallyliketothankDr.LinandDr.Leeforgivingmetheopportunityto workwiththemandguidemeattheCenterforUrbanTransportationResearchCUTRatUSFin theITS,TrafcOperationsandSafetyProgram.Thedatausedinthisdissertationwascollected duringaresearchprojectsponsoredbytheFederalHighwayAdministrationandmanagedbythe FloridaDepartmentofTransportationDistrict7.IwouldalsoliketothanktheGraduateResearch AssistantsattheITS,OperationsandSafetyProgramatCUTRfortheirassistanceinelddata collectionanddatareduction.

PAGE 4

NotetoReader:Theoriginalofthisdocumentcontainscolorthatis necessaryforunderstandingthedata.Theoriginaldissertationisonle withtheUSFlibraryinTampa,Florida.

PAGE 5

TableofContents ListofTables iv ListofFiguresvii ListofAcronymsx Abstract xi Chapter1Introduction1 1.1Background1 1.2ProblemStatement3 1.3ResearchObjectives3 1.4OutlineofDissertation4 Chapter2LiteratureReview5 2.1BackingCrashProblemSizeandDescription5 2.1.1BackingCrashTypes6 2.1.2UnderControlBackingCrashes7 2.1.3VehicleTypes7 2.1.4BackingCrashFatalities8 2.1.5BackoverCrashCharacteristics9 2.2BackoverCrashesIncludingCommercialTrucks10 2.3BackingCrashCountermeasures11 2.3.1SensorBasedCountermeasures12 2.3.2VideoBasedCountermeasures13 2.3.2.1RearviewCameras15 2.3.2.2Monitors17 2.4PreviousTestingEffortsonSimilarorSameDevicesinaControlled Environment18 Chapter3ResearchMethodology22 3.1CrashAnalysisReportingSystem22 3.2ExperimentDesign24 3.2.1StationaryObject25 i

PAGE 6

3.2.2SurprisePedestrianSurrogate25 3.2.3TestingProcedure26 3.2.4DataReduction27 3.3StatisticalModelDevelopment28 Chapter4DataCollectionMethodology31 4.1CrashDatafromCARSystem31 4.2ExperimentData32 4.2.1DataAcquisitionSystem32 4.2.2BackingManeuvers33 4.2.2.1StraightLineBackingManeuver35 4.2.2.2OffsetRightBackingManeuver38 4.2.2.3AlleyDockBackingManeuver40 4.2.3DeterminationofExperimentVariables41 Chapter5CrashDatabaseAnalysis44 5.1BackingCrashes44 5.2CrashReportsAnalysisResults49 5.2.1NoInjuryBackingCrashes50 5.2.2PossibleInjuryBackingCrashes53 5.2.3Non-IncapacitatingInjuryBackingCrashes56 5.2.4IncapacitatingInjuryBackingCrashes57 5.2.5FatalBackingCrashes59 5.3IdenticationofApplicableBackingCrashes61 Chapter6DriverBehaviorAnalysis64 6.1DriverVisualAttention66 6.2DriverAcceptance83 6.3AnalysisResults91 6.3.1DriverVisualAttentionResults91 6.3.2DriverAcceptanceResults92 6.3.3AccuracyTestResults93 Chapter7GeneralImplementationforCamera/Video-BasedDevices96 7.1Scientic/EngineeringGuidelines96 7.2UserAcceptanceAssessmentMethodology98 7.3DeviceUsagePatterns101 Chapter8ConclusionsandFutureResearch102 8.1Summary102 8.2ConclusionsandRecommendations105 8.3FutureResearch106 ListofReferences107 ii

PAGE 7

Appendices 110 AppendixA:RearviewSystemComponentTechnicalSpecications111 AppendixB:InformedConsentForm113 AppendixC:AfterExperimentEvaluationForm117 AbouttheAuthorEndPage iii

PAGE 8

ListofTables Table2.1BackoverfatalitiesinFARSbyyear[1].8 Table2.2Backoverdeathsidentiedin1998deathcerticatesbylocation[2].9 Table2.3BackoverdeathsidentiedinFARS-2001[2].9 Table2.4Rateofon-roadfatalbackingcrashes.12 Table2.5Studiesrelatedtorearobjectdetectiontechnologies.19 Table3.1Totalcrashes,Florida-2006.23 Table4.1Variablescreatedfordataanalysis.41 Table5.1Percentofbackingcrashestototalcrashes,Florida-2006.45 Table5.2Sitelocationofbackingcrashes,Florida-2006.48 Table5.3Backingcrashinjuryseverity,Florida-2006.48 Table5.4CrashlocationforNoInjurybackingcrashes.50 Table5.5EstimatedspeedofvehiclesforNoInjurybackingcrashes.51 Table5.6VehicletypesinvolvedinNoInjurybackingcrashes.53 Table5.7CrashlocationforPossibleInjurybackingcrashes.54 Table5.8EstimatedspeedofvehiclesforPossibleInjurybackingcrashes.55 Table5.9VehicletypesinvolvedinPossibleInjurybackingcrashes.55 Table5.10CrashlocationforNonIncapacitatingInjurybackingcrashes.56 iv

PAGE 9

Table5.11EstimatedspeedofvehiclesforNon-IncapacitatingInjurybackingcrashes.57 Table5.12VehicletypesinvolvedinNon-IncapacitatingInjurybackingcrashes.58 Table5.13EstimatedspeedofvehiclesforIncapacitatingInjurybackingcrashes.58 Table5.14VehicletypesinvolvedinIncapacitatingInjurybackingcrashes.60 Table5.15EstimatedspeedofvehiclesforFatalbackingcrashes.61 Table5.16VehicletypesinvolvedinFatalbackingcrashes.61 Table5.17Pointofimpactofat-faultvehicleduringallbackingcrashes.62 Table5.18Speedofvehiclesattimeofallbackingcrashes.62 Table6.1Averagemaneuvertimes.68 Table6.2Summarizedoutcomes.70 Table6.32X2Contingencytablesforthethreetests.70 Table6.4ResultofMcNemartest.71 Table6.5Coefcientsforrstmodel.71 Table6.6MonitorglancefrequencyANOVAtesting.73 Table6.7MonitorglancedurationANOVAtesting.74 Table6.8Variablesusedforbehavioranalysis.76 Table6.9Correlationmatrixforthe8driverattentionvariables.77 Table6.10Totalvarianceexplained.77 Table6.11Rotatedcomponentmatrix.79 Table6.12Principalcomponentscorrelationmatrix.79 Table6.13Variablesinthemodel.81 Table6.14Classicationstatisticsfordifferentdecisionrules.81 Table6.15Pairedsamplesdescriptivestatistics.94 v

PAGE 10

Table6.16Pairedsamplescorrelations.95 Table6.17Pairedsamplestest.95 Table6.18Distancetodummyanddockformaneuvers.95 TableA.1 1 3 inchCCDrearviewcameraspecications.111 TableA.2LCD5inchmonitorspecications.112 vi

PAGE 11

ListofFigures Figure2.1LargetruckrearNoZone,cameralocationandcameraeldofview.15 Figure2.2Viewfromrearviewcameralocatedat3ftfromgroundlevel.15 Figure2.3Viewfromrearviewcameralocatedat13ftfromgroundlevel.16 Figure2.4LCD5inchdisplaymountedondashboardoflargetruck.16 Figure2.5 1 3 inchrearviewcamerawithinfraredLEDs.17 Figure3.1Pedestriandummybehindtruck.26 Figure3.2Pedestriandummyclose-up.26 Figure3.3Video/Sensoranalysissoftwarescreen.28 Figure3.4Driverfourglancelocationsusedfordataanalysis.29 Figure4.1DAS-Outsideview.33 Figure4.2DAS-Insideview.33 Figure4.3TheDAScasewiringdiagram.33 Figure4.4Locationofthethreecamerasusedintheexperiment.34 Figure4.5Theexperimentdataprocessing/storageprocess.35 Figure4.6DiagramofStraightLineBackingManeuver.37 Figure4.7StraightLineBackingManeuverconesequence.38 Figure4.8DiagramofOffsetRightBackingManeuver.39 vii

PAGE 12

Figure4.9Pedestriandummyintroduction.40 Figure4.10DiagramofAlleyDockBackingManeuver.41 Figure5.1Pointofimpactofat-faultvehicleinbackingcrashes.47 Figure5.2ThemostcommoncrashlocationsforNoInjurybackingcrashes.52 Figure5.3ThemostcommoncrashlocationsforPossibleInjurybackingcrashes.54 Figure5.4ThemostcommoncrashlocationsforNon-IncapacitatingInjurybacking crashes.57 Figure5.5ThelocationcategoriesforIncapacitatingInjurybackingcrashes.59 Figure5.6ThelocationcategoriesofFatalbackingcrashes.60 Figure6.1Thedriveragedistribution.64 Figure6.2Drivers'CDLdrivingexperiencedistribution.65 Figure6.3Thebackingcrashhistoryofthesampleinnumberofcrashes.66 Figure6.4Driversidelocation.67 Figure6.5Passengersidelocation.67 Figure6.6Monitorlocation.67 Figure6.7Otherlocation.67 Figure6.8Themaneuvertimeline.68 Figure6.9TheglancepatternofadriverperformingtheStraightLineBackingManeuver.69 Figure6.10TheSCREEplotprovidedbySPSS.78 Figure6.11Graphshowingrepliestoquestion:WiththeRVSIfeelmorecomfortable performingtheStraightLineBackingManeuver.83 Figure6.12Graphshowingrepliestoquestion:WiththeRVSIfeelmorecomfortable performingtheOffsetRightBackingManeuver.84 viii

PAGE 13

Figure6.13Graphshowingrepliestoquestion:WiththeRVSIfeelmorecomfortable performingtheAlleyDockBackingManeuver.85 Figure6.14Graphshowingrepliestoquestion:TheRVShelpstoreducepotentialcrashes duringtheStraightLineBackingManeuver.86 Figure6.15Graphshowingrepliestoquestion:TheRVShelpstoreducepotentialcrashes duringtheOffsetRightBackingManeuver.87 Figure6.16Graphshowingrepliestoquestion:TheRVShelpstoreducepotentialcrashes duringtheAlleyDockBackingManeuver.88 Figure6.17Graphshowingrepliestoquestion:WouldyouliketohavetheRVSonthe truckyoudriveeveryday.89 Figure6.18Graphshowingrepliestoquestion:TheRVShelpstoreducetherearblind spotofthevehicle.90 Figure6.19Drivermeanglancepersecondduringexperiment.91 Figure6.20Drivermeantimeperglance.92 Figure6.21Backingcrashesexperiencedduringexperiment.94 ix

PAGE 14

ListofAcronyms ANPRMAdvancedNoticeforProposedRulemaking CARFlorida'sCrashAnalysisReportingSystem CDLCommercialDriverLicense CRTCathodeRayTube CUTRCenterforUrbanTransportationResearch FARSFatalityAnalysisReportingSystem FMCSAFederalMotorCarrierSafetyAdministration FMVSSFederalMotorVehicleSafetyStandards GESGeneralEstimatesSystem GVWRGrossVehicleWeightRating LCDLiquidCrystalDisplay LEDLightEmittingDiode NHTSANationalHighwayTrafcSafetyAdministration OEMOriginalEquipmentManufacturer PDOPropertyDamageOnly RVSRearviewVideoSystem SAFETEA-LUSafe,Accountable,Flexible,EfcientTransportationEquityAct:ALegacy forUsers TRIPTheRoadInformationProgram VMTVehicleMilesTraveled x

PAGE 15

OperationalEvaluationofAdvancedSafetyEnhancementDevices:RearviewVideoSystem AchilleasKourtellis ABSTRACT Sincethecreationoftheautomobile,therehasbeenanefforttocreateandimplementmechanical andelectronicdevicesthatwouldimprovevehiclesafety.Inrecentyears,electronictechnologies havebecomemoreefcientandcosteffective,thereforecreatingagreatspikeinwidespreadimplementation.Thesesafetyrelateddeviceshavetobetestedfortheirreliabilityandamountofhelp theyprovidethedriverwith.Theenduserthedriverhastobeinvolvedforasuccessfuldevice. Thisresearchpresentsthemethodologyusedtoevaluatetheeffectivenessoftherearviewvideo systemRVSusedinvehicles,especiallyinlargecommercialtrucksandeffectivelythemethodologyforamorecompleteinvestigationoftheproblemofcorrectlyimplementingasafetydevice. Thefocusofthisresearchisbackingcrashesthatinvolvelargetrucks.Thecountermeasuretested wasarearviewvideosystemwhichprovidesarearviewtothedriverinrealtime.Atraditional crashdataanalysisisalmostimpossiblesincethereisnotenoughdatatoperformit,andnodata areavailablefortheuseofthissystemsinceitisfairlynewtothemarket.Adriverexperiment undercontrolledconditionswasusedtocreateandcollectthedatanecessaryfortheanalysis.The experimentyieldedatotalof71crashesoutof270maneuvers.3%.Whenanalyzed,three backingneuversyieldeddifferentprobabilitiesofhavingabackingcrashwithandwithoutthe RVS.Theincreaseinstopraterangedfrom46.67percentto4.44percent.Thisisinterpretedas crashreductionduetothedevice.Driverbehaviorwasobservedduringtheexperimentandmeasuredforsignicantdifferences.Thedriversneededonaverage6.47secondsmoretimeforthe maneuverswiththeRVSinuse.Theyspentlesstimelookingatmirrorsanddiditlessfrequently inordertoaccommodatetheadditionalglancelocationpresentedtothem.Overalltheyseemedto beabletomanagetheirtimewithsomeexceptions. xi

PAGE 16

Thedriveracceptanceofthedevicewasalsomeasuredwithasurveygiventothemafterthey completedthetest.Overallinallmeasuresthemajorityofdriversagreedthatthesystemhelps inreducingtherearblindspotandthusitisahelpfuldeviceinreducingbackingcrashessinceit willhelpthemavoidpotentialhazardswhilebacking.Themajorityalsostatedthattheywould liketohavethedeviceintheirtruckforeverydayoperations.Theseresultsshowanacceptance ofthedeviceandthereforethemaximizationofthedevice'suseandpotentialbenets.TheRVS isthereforeeffectiveinreducingpotentialbackingcrashes.Theresultspresentedherearelimited, andinferencesaremadewiththeexperimentconditionsinmind.Generalapplicationoftheresults ispossible,withcertainassumptionsandrestrictions. xii

PAGE 17

Chapter1:Introduction In-vehiclesafetydeviceshavebeenincreasinglyusedintheindustrytoenhancethesafetyofthe driversandpassengersofvehicles.Sincethedevelopmentofthesedevicesisprimarilydrivenby research,acertainamountofsuchsafetyenhancementscomesintheaftermarketandnotinthe OriginalEquipmentManufacturerOEMindustry.Thishowever,createstheproblemofadditionaldevicesbeingpresentindrivers'cabinsotherthanoriginalequipment,especiallyinthecase ofcommercialvehicles.Inadditiontosafetydevices,infotainmenttelematicdevicesarebecoming increasinglypopular,thusincreasingthedrivers'stimulusandglancelocations.Inrecentyears moresuchdeviceshavebecomeavailableintheaftermarket,thusprovidingindividualdriversor companieswitheetstheopportunitytoequiptheirvehicleswiththelatesttechnologybutfora fractionofthemanufacturer'sprice. Whenamanufacturercreatesadeviceafterrigorousresearchanddevelopmentprocess,thedevice mustbetestedforitsreliability,andeffectiveness.Aproblemoccurshoweverwhenthedevice isdevelopedandimplementedwithouttestingorfeedbackfromtheendusers,inthiscasethe drivers.Itwasfoundthatagainstbelief,adevicethatshouldbeveryusefultodriversisnotused becauseitwassimplynotaccepted.Apropermethodologyencompassingdriverusebehaviorand feedbackneedstobeimplementedtosecureamorecompleteapproachindevelopingortesting suchdevices. 1.1Background Ingeneral,theoperationoflargetrucksinvolvesmanydifferenttypesofmaneuvers.Thebacking maneuver,inparticular,requiresahigherlevelofdriverattentionduetothelimitedview.Most 1

PAGE 18

truckingcompanieshaveapolicythatencouragesdriverstovisuallychecktherearofavehicle beforeperformingabackingmaneuver,regardlessofthebackingdistance,andtouseaspotter whocanstandoutsidethevehicletoensuresafetyduringthemaneuver. AspartofthefederaltransportationauthorizationknownastheSafe,Accountable,Flexible,EfcientTransportationEquityAct:ALegacyforUsersSAFETEA-LU,theU.S.Congresshas identiedahigh-priorityneedforresearchtoreducefatalities,injuries,andpropertydamagecaused bybackingcrashesinvolvingtrucks.Thesebackingcrashesareoftencausedbythepresenceof largeblindspots,commonlyknownastheNoZone,wherethetruckdriverhasvirtuallyno visibility. Blindspotsorareasinthecontextofdrivinganautomobilearethoseareasoftheroadthatcannotbeseenwhilelookingforwardorthrougheitherrearvieworsidemirrors.Blindspotscanbe eliminatedbyoverlappingsideandrearviewmirrors,bythedriverphysicallyturningaroundto lookbackwards,orbyaddinganothermirrorwithalargereldofview.Detectionofvehicles orotherobjectsinblindspotsmayalsobeaidedbysystemssuchasvideocamerasordistance sensors.Blindspotscanbeatanylocationaroundavehicle,dependingonthesizeandstructure ofthevehicle,orpresenceofvehiclefeaturessuchasA-pillars.Therefore,therecanbeside,front, andrearblindspots. Rearblindspotsareofthemostconcerninlargetrucksorcommercialvehicles.Usually,these vehiclesdonothavearearwindoworarearviewmirror,whicharetheprimarymethodsforeliminatingblindspots.Inaddition,thelargesizeoftrucksandcommercialvehiclesmakestherear blindspotadangerousareaorzone.Ifasmallervehicleorpedestrianisinthisareawhilethe truckisbacking,thereisagreatpotentialforacrashsincethetruckdrivercannotseethem.This rearblindspotincreasesasthevehiclesizeincreases. Itisbelievedthatthereisaverylargepotentialtoreducebackingcrashesbyreducingoreliminatingtherearblindspotusingrearviewvideotechnologiesthatmaketheblindspotvisibleto thedriverthroughavideosystem.Thisstudyaimstoevaluatetheeffectivenessofrearviewvideo systemsasacountermeasureforlargetruckstoreducepotentialbackingcrashes.Alargetruck isdenedasatruckwithagrossvehicleweightratingGVWRgreaterthan10,000poundsand 2

PAGE 19

includesbothmediumandheavytrucks.Usually,thesetrucksareatractor-trailercombination, althoughcertaincategoriesofheavytrucksareasingleunit. 1.2ProblemStatement Sincethecreationoftheautomobile,manhasstrivedtocreateandimplementmechanicaland electronicdevicesthatwouldimprovevehiclesafety.Especiallyinmorerecentyears,whenelectronictechnologieshavebecomemoreefcientandcosteffective,thereforecreatingagreatspike inwidespreadimplementation.Whenmanufacturerscreatedevices,especiallyhumaninterface devices,theyhavetomakesurethatthedevicewillhelpthedriverwithoutincreasingthemental orphysicalworkloadthushinderingthesamesafetytheyweremeantfor.Ithasbeenshownthat increasingthedriverworkloadcancauseseveredeteriorationofthedriver'sattentionandabilityto respondfasttoadangeroussituationcausingacrash.Theprocessinvolvedinthedevelopment ofsuchdevicesinvolvesidenticationofaproblem,andimplementationofideastosolvethe problem.Oftenthedeviceismadeandinstalledonvehicleswithouthavingthedirectinvolvement oftheenduser:thedriver.Amoreholisticapproachmustinvolveidenticationoftheproblem, developmentofthedeviceandtestrunswithawidepanelofdriverstogaininsightastotheusageofthedevicetobeabletoimplementitmoreeffectively.Especiallydevicesthatrequirethe driver'sattentionpassivedevices,suchasmosttelematicdevices,havetobeimplementedinthe bestwaypossibletobeusedbythedriver. 1.3ResearchObjectives Thisresearchinvolvedseveralobjectivestoencompassinrealizingthemainobjectiveofevaluatingtheoperationaleffectivenessoftherearviewvideosysteminlargetrucks.Theobjectivesare describedas: Analyzecrashdatatoobtainunderstandingontheproblemsizeandconditions, Identifyamountandcharacteristicsofbackingcrashesthesystemisapplicableto, 3

PAGE 20

DesigndriverexperimentincontrolledconditionstotesttheeffectivenessoftheRVSin reducingbackingcrashes, Analyzedatafromtheexperimentandinferonif,whenandhowthesystemhelpsthedriver, Provideanalysisonthedriverbehaviorandusageofthesystemandrecommendationson thedevice'susage. 1.4OutlineofDissertation Thisdocumentconsistsofeightchapters.Chapter1providesanintroductiontotheresearchproblemandoutlinestheresearchobjectives.Chapter2summarizesaliteraturereviewinthisarea. Chapter3describestheresearchmethodologiesutilizedtoreachtheobjectivesofthestudy.Chapter4describestheproceduresfollowedtocompletedatacollectioninanefcientandappropriate manner.Chapter5includesanalysisresultsandndingsfromacrashdatabaseonbackingcrashes. Chapter6summarizestheresultsofdataanalysisfordriverbehaviorduringtheexperiment.Chapter7providesadiscussionimplementationofthemodelstoamoregeneralproblemordevice. Finally,chapter8providessummary,conclusionsandrecommendationsfromthisresearch,aswell asdiscussiononfutureresearch. 4

PAGE 21

Chapter2:LiteratureReview Abackingcrashoccurswhenabackingvehiclestrikesanothervehicle,stationaryobject,abicyclist,amotorcyclistorapedestrian.ConsiderableresearchhasbeenperformedbytheNational HighwayTrafcSafetyAdministrationNHTSAoverthepasttwodecadestoidentifyhowback crasheshappenandevaluateavailablecountermeasurestopreventandreducethem.Themain causeofthesecrashesistherearblindzonethedriverexhibitsdirectlybehindthevehicle.This zonebecomeslargerwithincreasingvehiclesizeandlength.Itisespeciallydangerousinthecase oflargetrucks,eithersingleormulti-unitduetothesizeofthevehicleandthefactthatthedrivers donothaveareareldofviewavailable. Backingcrashesareasmallpercentageofalltypesofcrashes.Studiesconcerningbackingcrashes havebeenconductedthroughthelast20years.Onemajorcharacteristicofbackingcrasheshoweverthatdoesnotapplytoallothercrashesisthattheycanbeavoidedbyrationaldriverbehavior andarearvisibilitythatcoverstheareabehindanyvehicle. 2.1BackingCrashProblemSizeandDescription Backingcrashesoccuratasmallratecomparedtoothercrashes.Obviouslythisisduetothesmall amountofexposurethisparticularhasincomparisontoallothercrashtypes.Theyarenotnegligiblenonetheless.Amajorproblemwithbackingcrashreportingisthefactthatalargemajorityof thesecrasheshappenonprivateproperty.Thiswillbeshownlaterinbackingcrashesanalyzed fromaFloridacrashdatabasefortheyears2003-2006.Sinceallcrashdatabasesincludeonly crashesthatoccurinpublicroads,thecrashesthatoccuronprivatepropertyarenotreported.This waytheactualpercentageofbackingcrashescannotbeknown,butestimationmethodscanbe 5

PAGE 22

utilizedtoobtainabetterunderstandingoftheproblemsize.Fortunatelyalargenumberofthe unreportedcrashesarepropertydamageonlybutsomeinjuriesandfatalitiesarealsoincluded. Therststudyfoundtoexaminebackingcrashes[3],waspublishedin1994with1990data.The teamidentiedtheproblemofbackingcrashesin1990tobe181,500policereportedbacking crasheswith185associatedfatalities.Thesenumbersdonotincludeoff-roadwaycrashessuch asdrivewaybackingcrashes.Thestudyalsoreported: Approximately22,000associatedinjuries,including1,500seriousinjuries. Approximately91percentofallbackingcrasheswerePropertyDamageOnlyPDO. Backingcrasheswere3percentofallcrashesbutaccountedforonly0.4percentofallfatalities. Duringitslifetimeavehicleisexpectedtobeinvolvedin0.01policereportedbackingcrashes asthebackingvehicle. Approximately300,000non-policereportedbackingcrasheswereestimatedfor1990. Backingcrashescauseabout1percentofallcrash-causeddelay. Althoughthenumbershaveobviouslychangedthisshowsthattheproblemwasbigeventhen. Alsousefulanalysisshowedthatthecrasheswerecategorizedwithrespecttothevehiclemovements,thevehiclespeeds,andthevehicletypes.Thesecategoriesapplytodayaswell. 2.1.1BackingCrashTypes From[3],backingcrashesweredividedintotwomaintypes: 1.Slow-closingspeedencroachmentbackingcrashes aPedestrian/cyclist bParallelpath 6

PAGE 23

cCurvedpath 2.Crossingpathbackingcrashesinvolvinghigherclosingspeeds 2.1.2UnderControlBackingCrashes Furthermorethebackingcrashesweredividedintoundercontrolandnotundercontrolbackingcrashes.Theundercontrolbackingcrashesusuallyoccuratslowspeeds,andtheat-fault driverscanpotentiallyavoidthecrashiftheycanseetheothervehicle.Thenotundercontrol backingcrashesoccurwhenoneorbothdriverscannotavoidthecrashbecauseoffactorsthey donotcontrol.Disqualiedcasesfornotundercontrolbackingcrashesincluded:icy/snowy roadwaysurfaces,selectedvehicledefects,driverlessvehicles,grossly-intoxicateddriversand selecteddriverphysicalimpairments.Thesecasesweredeemednotundercontrolbecausethe driverwouldpotentiallynotbecapabletorespondtoacollisionwarninggivenbyacollision warningsystemcountermeasurethusthebackingcrashwouldbeunavoidable.Themainreason foraggregatingthecrashesintothesecategorieswastoestablishwhichcrashescouldbepotentiallyavoidedwithacountermeasurenamelybackupsensorsorrearviewsystems.AstheRVS becomesmorepopular,therehavebeengrowinginterestsonvariousaspectsofRVSincludingthe benet/costofusingtheRVS. 2.1.3VehicleTypes Thevehicletypesinvolvedinbackingcrashesare: Passengervehicles:cars,lighttrucks,andvans, Mediumsizetrucks:single-unittrucks, Largesizetrucks:singleunitorcombinationunittrucks. 7

PAGE 24

2.1.4BackingCrashFatalities Fatalitiesthatoccurfromabackingcrasharecalledbackoverfatalitiessincethevehicleusually backsoverthepersonbeingapedestrian,acyclistoramotorcyclist.Arecentstudyexaminedthe fatalitiesofbackingcrashesforallvehicles[1],andestimatedthatthereareabout183fatalities frombackingcrashesperyear.Table2.1showsthenumberoffatalitiesoccurringfrombacking crashes.Theestimationof183fatalitiesperyearwasbasedondeathcerticatedata,Fatality AnalysisReportingSystemFARSdata,andGeneralEstimatesSystemGESdata. Table2.1:BackoverfatalitiesinFARSbyyear[1]. YearFatalities 199183 199280 199376 199484 199575 199679 199759 199883 199970 200089 200166 200266 200379 200472 Thefatalities,however,areonlyasmallportionofallthebackingcrashes.CrashdatafromFlorida's CrashAnalysisReportingSystemCARshowthatonaveragetherearemorethan14,000police reportedbackingcrashesperyearinFlorida.Thisaccountsforabout4.61percentofallcrash types.Althoughthisnumberissmall,thedrivebehindthisresearchisthatalargenumberofthis crashescanbeavoidedusingtechnologycountermeasures.Aswillalsobeshownlater,largetrucks haveamuchlarger-almostdouble-percentageofbackingcrashesthanpassengervehicles. Inadditiontothefatalitiesfrombackingcrashes,areporttoU.S.Congressin2006showedthat therearebetween6,700and7,419injuriesfrombackingcrashesperyear[2]. 8

PAGE 25

2.1.5BackoverCrashCharacteristics Accordingto[2],inanalysisofdeathcerticates,14percentofbackoverdeathswerefoundthat couldbecalledtheRoad/Streetoccurrences.Thespeciclocationsofthesecrashesareshown inTable2.2.Thistablepresentscategoriesbasedonthelimiteddetailsprovidedonthedeath certicatesandinsomecasessupplementaryinformatione.g.newspaperreports.Thesedata suggestthatcasesoccurpredominantlyawayfromstreetsandroads.AreviewofFARScases from2000and2001bylocationalsorevealsthatbackoverfatalitiesoccurpredominantlyaway fromtheroadsorstreets,showninTable2.3. Table2.2:Backoverdeathsidentiedin1998deathcerticatesbylocation[2]. LocationFrequency*Percent Driveway2123 Home2123 ParkingLot2123 Road/Street1314 Sidewalk22 OtherOffRoad1314 *Dataarefrom35StatesandtheDistrictofColumbia. Table2.3:BackoverdeathsidentiedinFARS-2001[2]. LocationFrequencyPercent Driveway4443 ParkingLot55 Road/Street2827 OtherOffRoad2525 Throughthesestudies,itwasalsofoundthatbackingcrashesoccurlargelyduringdaytimewithno adverseweatherconditionsorothermajorenvironmentalcontributingfactors.Therewasahigher involvementratepermillionVehicleMilesTraveledMVMTforyounger-19andolder +drivers,aswellasmaleshada7.5per100MVMTtofemales4.3per100MVMT.Amajor causeforbackingcrasheswasrecognitionerror/improperlookout.Thedriverofthebacking 9

PAGE 26

vehicleeitherfailedtolookorlookedbutdidn'tsee[3].Althoughregisteredvehiclesincrease throughtime,andonewouldexpectanincreaseintherateofbackingcrashes,itseemsthatthe trendisrelativelyconstant.Accordingto[4],Despitetheabovetrendsintheeetwithrespectto vehicleregistrationsandincreasingvehiclesize,FARSbackoverfatalitiesanalyzedforthisreport -2004donotshowanincreasingtrend.Infact,thereisanon-signicant,yetdecreasing trend.However,therearenoaccuratetrenddataspecicallyforthenon-trafcincidentsthatmay ormaynotfollowthefatalitytrendsseeninthetrafccrashdatabases.ALexisNexisreviewof periodicalsonbackoverincidentsincludingnon-trafccrashesfrom1998to2002wasalsonot abletodemonstrateacleartrend. 2.2BackoverCrashesIncludingCommercialTrucks Over15billiontonsofgoods,worthover$92trillion,aremovedannuallyintheU.S.-theequivalentof310poundsoffreightbeingmoveddailyforeachU.S.resident.Thelargestshareofthe nation'sfreightismovedbytrucks,whichcarry71percentofalltonnageand80percentofthe valueofU.S.shipments,asignicantfactortotheU.S.economy[5]. Dataonbackovercrashesacrossvariousstatesinthenationwerecollectedandusedtoassessthe impactofreducingthesecrashes.Informationonbackovercrasheswasresearchedinpublications, printedarticles,journals,onlinedatabases,andwebsitesofcompaniesthatprovideproductsfor reducingbackovercrashes.Statisticsonbackovercrasheswereavailableinsomepublicationsand onlinedatabasesonsafetyandsecurity. TrafcaccidentsinvolvinglargetruckswithGrossVehicleWeightRatingGVWRover10,000 poundsareresponsibleforasignicantproportionoftrafcfatalitiesannually,andaccidents involvinglargetrucksaremorelikelytoresultinfatalities,duetothemoreseriousconsequences ofaccidentsinvolvinglargervehicles.TheRoadInformationProgramTRIPanalyzeddatafrom theNationalHighwayTrafcSafetyAdministrationNHTSAonfataltrafcaccidentsinvolving largetrucksintheU.S.from1998to2002togainabetterunderstandingofthecharacteristicsof theseaccidents[6].Itwasfoundthat26,065ofthe210,174trafcfatalitiesthatoccurredfrom 10

PAGE 27

1998to2002intheU.S.-approximatelyoneofeight-resultedfromacollisionthatinvolveda largetruck.Fataltrafcaccidentsinvolvinglargetrucksfrom1998to2002resultedinthedeaths of3,647personswhoweredriversoroccupantsoflargetrucks,andtheremaining22,418people killedwereeitherdriversoroccupantsofothervehiclesorwerenon-motorists,suchaspedestrians orcyclists.Thus,approximatelysixoutofsevenpeoplekilledinfataltrafcaccidentsinvolving largetruckswerenotoccupantsofalargetruck. Ithasbeenreportedthatabout67percentoffatalcrashesduetobackingofheavytrucksgounreportedincrashdatabasesbecausetheydonotusuallyhappenonroads[2].Accordingtothe FloridaTrafcCrashStatisticsReport2005[7],improperbackingwasthecauseof218crashes ofthetotal5,709crashesinvolvingaheavytruck.TheStateofMichiganreportedatotalof777 heavytrucksorbusesinvolvedinimproperbackingcrashesofthetotal16,238heavytruckor busesinvolvedinacrashintheyear2005[8].During2005inMinnesota,4.4percentofthetotal 4,150contributingfactorscitedinlargetruckcrasheswereunsafebacking[9].Accordingtothe 2004WisconsinTrafcCrashFacts,394driver-relatedcircumstanceswereattributedtounsafe backingbythedriverofalargetruckin7,898largetruckcrashes[10].In2004,backingresulted in639crashesofthetotal13,908crashesinvolvingalargetruckorbusonallstate-maintained roadsinNorthCarolina[11].Improperbackingwastheprimarycollisionfactorin107fatalor injurycrashesofthetotal3,762fatalorinjurycrashesinvolvingatruck[12].TheNewYorkState DepartmentofMotorVehiclesreported110crashesofthetotal5,410largetruckcrashesasa resultofunsafebackingin2004.AsshowninTable2.4,theratiooffatalitiesoflargetrucksfor backingpermillionregisteredvehiclesandper100billionvehiclemilestraveledissignicantly largerthananyothertypeofvehicles. 2.3BackingCrashCountermeasures Inthelast15years,therehasbeendevelopmentoftechnologicaldevicesthataddresstheissueof therearblindspotinvehicles.Ithasbeenidentiedbyresearchstudiesthatthemajorcauseof 11

PAGE 28

Table2.4:Rateofon-roadfatalbackingcrashes. CumulativeFARSdatafrom1991-1997 PedestriansandCyclistsKilledPedestriansandCyclistsKilled byaBackingVehicleperbyaBackingVehicleper100 VehicleTypeMillionRegisteredVehiclesBillionVehicleMilesTraveled Passengercars1.051.26 Lighttrucks/vans2.322.80 Combinationtrucks9.942.21 Straighttrucks29.6821.89 backingcrashesisthelimitedornorearvisibility,thusdevicesthathelpthedriveridentifyobjects behindthevehicleshouldhelpinavoidingpotentialbackingcrashes. Abackingcrashcountermeasureshouldprovidethedriverwithinformationonobjectsbehindthe vehiclewhilebacking.Itisreasonabletoassumethatifthedriverknowsthatthereisanobjector personbehindthevehiclewhiletheyarebacking,thedriverwillstopthevehicletoavoidstriking theobjectorperson.Inthecaseoflargetrucks,thereisnorearviewmirrorsincethereisnorear window,thusthedriverhastorelysolelyonthetwosideviewmirrorsforbackingmaneuvers. Thecountermeasurescanbedividedintotwomaincategories: 1.Sensorbasedtechnologies 2.Videobasedtechnologies Anothervisualcountermeasurepreviouslyusedwasconvexmirrorscross-viewmirrorsmounted inthebackupperleftcornerofthetruck.Thedriverwasabletoidentifylargeobjectsbehind thetruckusingthedriversidemirrorandthecross-viewmirror.Thiscountermeasurewillnotbe discussedsinceitisrelativelyoldandunreliable. 2.3.1SensorBasedCountermeasures Thesecountermeasuresusuallyinvolveultrasonicorradarsensorsmountedonthebumperofthe vehicle.Thesensorsemitwaves,andifanobjectispresentincloseproximitytothevehicle,the reectedwavesarecapturedandtranslatedtoavisualoraudiblewarningsothedriverknowsthat 12

PAGE 29

thereissomethinginthepathofthebackingvehicle.Differentsensorshavedifferentrangesand capabilities.Thesesystemswererstdevelopedtobeusedasparkingaids,sothatthedrivercan parkthevehicleduringslowparkingmaneuverswithaccuracyandsafety.Thesesystemshave beenwidelyusedinthepastonpassengercars,vansandlighttrucks.Theyareusuallycalled backupsensorsandarealsoavailableforcommercialvehicles. ResearchstudiesconductedbytheNHTSAhaveshownhoweverthatthesesystems,arenotaccurateenoughtobetrustedwiththetaskofpersondetection,themostseveresituationofabacking crash.Sincetheearly1990'sthesesystemshavebeenavailabletobeusedonvehiclesforthese purposes.Therststudiesshowedapotentialforthesensorsystemstoaidthedriversavoidbackingcrashes[13].Alsonumerousstudieshaveinvestigatedthebenetsandlimitationsofusing suchsystems[14].Furtherexaminationhowever,showedthatthesesystemsarenotaccurate, haveinconsistentdetectionpatternsandlackreliabilityindetectingobjectsandpeoplebehindthe vehicles[18].Inreferencetocommercialtrucks,theNHTSAshowedthatthesystemscannot beusedasobjectdetectionsystems[2,21].ThelatestreporttotheU.S.Congresssummarized thefollowing: Findingsforultrasonicbackingsystems:Withrespecttothefunctionalgoalsofa backingsystem,neitherofthesetwosystemsmeetsanyoftherequirements.Evenfor nearzonedetectionbothsystemshaveamaximumrangeofabout9ft,notthe15ft calledfor.Ultrasonicbackingsystemswerefoundtobeextremelysensitiveandprone tofalsealarms.Backingsystemssufferfromorthogonalrequirements.Ontheone handonedoesnotwantthesystemtogooffallthetime,whileontheotherhandone wouldliketobesensitivetosmalltargets,suchaschildren,inanenvironmentwitha largeamountofgroundreturn. 2.3.2VideoBasedCountermeasures Inlateststudieswherebothsensorandvideosystemswerecompared,thevideosystemsperformedbetter.Theonlydifferencebetweenthevideosystemsandthesensorsystemsbeingthat 13

PAGE 30

theyarepassivesystems,thustheyrequirethedriver'sattentioninordertoprovidetheinformation.Thevideosystemscoverthewholerearareaofthevehicle,leavingthedriverbeingthe onlyfactorbetweendetectingornotdetectingtheobjects.Fromrecentstudiesperformedbythe NHTSA[2,21,22,24]thefollowingcanbesummarized: Findingsforrearviewvideosystems:Thevideosystemstestedappeartobequitecapableofextendingthedrivers'eldofregard.Thecontrastcompressionmayobscure sometargetsundercertainlightingconditions,butsuchaconditionwasnotobserved duringthesetests.Theeldofviewofbothsystemsprovidedadequatecoverage towardtherearofthevehicle.Thesesystemsarequitecapableofsatisfyingthetarget detectionfunctionalgoal.Obviously,theycannotsatisfythewarningrequirement. Sinceithasbeenshownthatthesensorsystemsarenotaseffective,onlyrearviewvideosystems weretakenintoconsiderationforthisresearch. RearviewvideosystemsRVSconsistofoneormorecamerasandamonitor.Therearecertain positionsinwhichthecamerascanbeplacedonatrucktoprovidearearview.Themainobjective istoprovidethedriverwitharearviewofthevehiclesotheycanuseitsimilartoarearviewmirrorinapassengercar.Figure2.1showstherearNoZoneofatruck,thelocationforacamera, andthecamera'seldofview.Usuallythecameraismountedattherearofthetruckatbumper heightftfromgroundleveloratthetopofthecargoboxofatruckftfromgroundlevelto eliminatetherearblindspot.Figure2.2andFigure2.3showviewsfromsuchcameras.Therear cameraisusedduringbackingmaneuvers,andtheviewcanbesetuptoautomaticallyswitchon whenthereversegearisengaged.Themonitorislocatedonthedashboardofthevehicleinthe lineofsightofthedriver.Figure2.4showsanexampleofadisplaymountedonthedashboardof asemitruck. 14

PAGE 31

Figure2.1:LargetruckrearNoZone,cameralocationandcameraeldofview. Figure2.2:Viewfromrearviewcameralocatedat3ftfromgroundlevel. 2.3.2.1RearviewCameras Thecamerascurrentlyusedinthesesystemsareusuallyencasedinawaterproofenclosureina robustmetalcasethatprovidesprotectioninextremeenvironmentalconditions.Sinceallcameras requireilluminationofthescenetoprovideanimage,thesecamerasareusuallyequippedwith infraredLightEmittingDiodesLEDs,whichhelpwhentheambientlightisnotenoughfor viewing,i.e.,atdusk,dawnorinshade.TheseinfraredLEDsilluminatetheareawithinfrared wavesthatarenotvisibletothehumaneye,butonlytothecamera,thusprovidinganimageeven incompletedarkness.Theimagehoweverisnolongerincolorbutinblackandwhitewhenthe nightvisionisused.TheinfraredLEDsareshowninFigure2.5,whichshowsthecameratype 15

PAGE 32

Figure2.3:Viewfromrearviewcameralocatedat13ftfromgroundlevel. Figure2.4:LCD5inchdisplaymountedondashboardoflargetruck. 16

PAGE 33

usedforthisstudywiththemountingbracket.Theparticularcamerahasa 1 3 inchlensandrequireszeroLuxilluminationsinceitusestheinfraredLEDsforlight.Thelensapertureisf/2.8. ThetechnicalspecicationsofthecamerausedinthisstudyareshowninTableA.1locatedin AppendixA.Figures2.2and2.3showviewsusingthesame 1 3 inchlenscamera. Figure2.5: 1 3 inchrearviewcamerawithinfraredLEDs. 2.3.2.2Monitors ThemonitorsusedinRVScanbeeithercathode-raytubeCRTorliquidcrystaldisplayLCD. CRTmonitorsarebulkyandrequirealargerspace,areheavierandolderintechnology,andcould bepotentiallymoredangerouswhenusedinvehiclesbecausethevacuuminsidethetubewould implode,andglassshatterwillcoverthedrivercabininjuringtheoccupantsinanaccident.The lightweightLCDmonitorsarethechoiceofmanyvehiclemanufacturersfordisplaysinsidethe vehiclesincetheyhavenoglasspartsundervacuum,arethinnerandlighterthanCRTs,andshow detailandcolorbetterthanCRTs.ThetechnicalspecicationsoftheLCDmonitorsusedforthis studyareshowninTableA.2locatedinAppendixA.Figure2.4showsa5inchLCDmonitor installedonthedashboardofavehicleasapartofarearviewvideosystem. 17

PAGE 34

2.4PreviousTestingEffortsonSimilarorSameDevicesinaControlledEnvironment PreviouseffortsonsimilartechnologiesincludeanumberofstudiesperformedbytheNHTSA, FMCSA,andotheragenciesforsensor-basedsystemsaswellascamera-basedsystems[2,17,21 23,26].Themajorityofthesestudiesperformedstatictestswiththesystemunderinvestigationinstalledonthevehiclewiththevehiclenotmoving,butobjectsorpedestriansmoving aroundthevehicletogaugeifthedetectionofthesystemwasadequate.Morerecentadvanced researcheffortsincludedriversimulatortests,wherethedriverswereaskedtodriveasimulated vehicleandrespondtocertainstimuli.Theobservationofthedriverswastheprimaryobjective oftheseefforts.Theonlystudythatperformeddynamictestsisdescribedin[26,28],wherethe driversdroveanexperimentaltruckequippedwiththesystemsunderevaluationandwereaskedto performcertainmaneuvers,withandwithoutthesystem,inseparateruns. Table2.5showsthehistoryofthesestudiesandthendingsofeach.Severalstudies,especiallyin the1990's,includedmirrorsinthelistoftechnologiesusedforbackingcrashavoidance.Current studiesfocusmoreonsensorandvideosystemsandcombinationsofbothtechnologies.Unfortunately,itisstilltooearlytodeterminehowwellthesesystemsworkonalargescale.Thelatest updatefromNHTSAisanAdvanceNoticeofProposedRuleMakingANPRMreleasedin2009 forFederalMotorVehicleSafetyStandardNo.111:RearviewMirrors,toimproveadriver'sabilitytoseeareastotherearofamotorvehicleinordertomitigatefatalitiesandinjuriesassociated withbackoverincidents[29].Thisnoticeasksfortheindustrytoprovidequestionsandcomments ontheproposedamendmentofrequiringadditionalsystemstoimprovethedriver'svisibilityin blindareas. 18

PAGE 35

Table2.5:Studiesrelatedtorearobjectdetectiontechnologies. YearAgencyTitleFindings 1994NHTSAAStudyofCommercialMotor VehicleElectronics-Based RearandSideObjectDetection Systems. Thisstudydealtwithmirrorandsensor systemsusedforsideandrearobject detectiononmarket-readyandprototype systems.Resultsshowedthattheyhave apotentialinhelpingdriversavoid crashes. 1995NHTSAHardwareEvaluationofHeavy TruckSideandRearObject DetectionSystems. Thisisthersttimeavideosystemis evaluatedforitspotentialsintruckcrash avoidancealongwithsensorsystems. Resultsshowedthatallsystemsimprove driverpotentialinavoidingcrashes. 2006NHTSAExperimentalEvaluationofthe PerformanceofAvailableBackoverPreventionTechnologies. Inthiscomprehensivestudy,mirror, sensor,andvideosystemswere testedfortheirperformanceonSUVtypevehicles.Resultsshowedthat videosystemshelpthedriverdetect pedestriansbehindtheirvehiclesbetter thanothersystems. 2007NHTSAEvaluationofPerformanceof AvailableBackoverPrevention TechnologiesforLightVehicles. Thisstudy,similartotheprevious,tested thesamesystemsonlightvehicles.The outcomewasthatthevideosystems performbetterinhelpingthedriver butdonotwarnthedriverofpotential danger. 19

PAGE 36

Table2.5:continued. YearAgencyTitleFindings 2007NHTSAExperimentalEvaluationofthe PerformanceofAvailableBackoverPreventionTechnologies forMediumStraightTrucks. Thesamestudywasperformedfor mediumtrucks.Theoutcomeshowed thatvideosystemsarethemostreliable toshowobjectsbehindatruck. 2007NHTSAUseofAdvancedIn-Vehicle TechnologybyYoungandOlder EarlyAdopters. Thisstudyshowedthedifferencesin agewiththeuseofsensorandcamera systems.Italsoshowedthatdrivers overestimatetheeffectivenessofthe systemsandaremorelikelytoback fasterandmorecarelesslywhenhaving thesystem. 2008NHTSADevelopmentofPerformance SpecicationforCamera-Video ImagingSystemsonHeavy Vehicle. Thisstudy,thelatestofitskind,tested camerasystemsfordifferentlocations onthetruck,toimprovevisibility andpotentiallyreplacemirrors.This studyincludeddynamictestingof systemswithdriversperformingbacking maneuvers. 20

PAGE 37

Table2.5:continued. YearAgencyTitleFindings 2008NHTSAOn-RoadStudyofDrivers'Use ofRearviewVideoSystems. Thelatestnaturalisticstudyofitskind collecteddriverbehaviordatafora monthfromminivandrivers.Theresults showedthattheeffectivenessofthe videosystemsisonlyabout20%and thedriversusethesystemsbutnotinan efcientmannertoavoidallpotential crashes. 21

PAGE 38

Chapter3:ResearchMethodology Thischapterdescribesthemethodologyfollowedinvarioussectionsofthisresearchinorderto achievetheproposedresults.Inordertomeasureandevaluatetherearviewvideosystem,agood understandingofthecharacteristicsofbackingcrashesmustbegainedinordertoeffectivelybe abletoapplythecountermeasure.Sincethissystemisdeemedacountermeasureforbacking crashes,thereasonsandcausesofbackingcrasheshavetoberstinvestigatedinordertotest thesystemagainstthoseissues.FirsttheFloridaCARSystemDatabasewasreviewedtoobtain moredetailsforthebackingcrashesinvolvinglargetrucks.Thenanexperimentwasdesignedto testdifferentaspectsofusingthesysteminrealsituations,andanalysisoftheresultsshowsthe importantfactorsthatinuencetheuseofsuchasystem. 3.1CrashAnalysisReportingSystem Florida'sCARdatabasewasusedtoobtaininformationonbackingcrashesinvolvinglargetrucks. Thedatausedwerefromyears2003-2006andincludedatotalof1,280,130crashes.Table3.1 showsthedatadistribution.Thepassengervehiclecategoryincludespassengercar,passengervan andlightpickuptrucktypes,whereasthelargetruckcategoryincludesmediumandheavytrucks andtrucktractors. Asshownlaterinchapter5,thedatawereanalyzedinreferencetothecausesofbackingcrashes. Thedatabasehasthreemajorvariablesrelatedtobacking: 1.Firstharmfulevent:backedinto 2.Vehiclemovement:backing 22

PAGE 39

Table3.1:Totalcrashes,Florida-2006. Year PassengerVeh.LargeTruckAllVehicleTypes Total%Total%Total 2003248,21582.4415,9755.31301,095 2004265,12481.9318,4605.70323,599 2005270,43681.3920,2836.10332,279 2006261,76781.0019,2135.95323,157 Total1,045,54273,9311,280,130 3.Contributingcause:improperbacking Therstvariableisthersteventcodedbythepoliceofcerasthersteventthatoccurredduring thecrash.Thisvariableisusedasthecrashtype.Thesecondvariable,isthevehicles'movement whenthecrashoccurredanditshouldbeobviouslybackingfortheat-faultvehicle.Thethird variableisthecontributingcauseascodedbythepoliceofcerafterthecrash.Thisisbasedon opinions,andstatementsofthedriversinvolvedandwitnesses.Althoughthesevariablesshouldbe consistent,asshownlaterthereisadiscrepancyinthenumbers. Basedonndingsfromsubsection2.1.1,thecountermeasurecannotbeappliedonalltypesof backingcrashes.Thecountermeasureisapplicableonlytocertainbackingcrashes.Therequirementsare: Bothvehiclesshouldtravelatlowspeeds<10mphbecausethevehicleshavetobeunder thecontrolofthedriversandgivethemenoughtimetorespondtoapotentialdanger, Theatfaultvehiclehasinitialpointofimpacttherearend,rearrightandrearleftcornersof thevehiclesincethesearethezonesthatarecoveredbytherearviewvideosystemandany rearobjectdetectionsystem. Allotherbackingcrashtypes,cannotbesaidtobepotentiallyavoidedsincesuchaclaimcannot besubstantiallysupported.Thecrashdataanalysisisdescribedindetailandresultsshownin chapter5. 23

PAGE 40

3.2ExperimentDesign Backingcrashesarearareevent.Aswithallcrashesthereisnowaytoknowwhenonewilloccur. Thatisthemainreasonsafetyanalysisisperformedoncrashdataaftertheevents.Inthiscase however,backingcrashesdonotoccurfrequentlyenoughsoinordertoaccumulateadequate samplesizeforstatisticalanalysis,manyyearsofdataareneeded. Asecondmethodforsafetyanalysisistheconictanalysis.Thismethodisnotthesameastrafc conictssincetheobservationisnotmadeintheroadundernormalconditions,butratherinacontrolledenvironment.Anexperimentcansimulatereallikeconditionsinacontrolledenvironment andusingvehiclesequippedwithdatacollectionequipmentitispossibletocollectadequatedata foranalysis. Theexperimentobjectiveswere: Toidentifyifthesystemhelpsthedriverperformabackingmaneuversaferthanwithoutthe system,i.e.tohelpthedriveravoidpotentialhazards, Toobservehowthedriversusethesysteminactualdrivingconditions, Toquantifymeasuresofeffectivenessinordertoevaluatethesystem, Toobtainfeedbackfromactualdriversforthesystem, Togainanunderstandingtothesystem'suse,functions,potentialsandlimitations. Theexperimentwasdesignedtousebackingmaneuverstotesttheeffectivenessoftherearview videosysteminaidingthedriverswithavoidingpotentialhazardsthatcouldleadtoacrash.One oftheproblemsrelatedwiththetestingofacrashisobviouslytosimulateacrashwithoutactually harminganyoneintheprocess.Asimplementedinpreviousstudiesincontrolledenvironments, pedestriansurrogateswereusedtosimulateactualpersonswalkingbehindthevehicles.Thedifferencewiththisexperimenthoweverwasthatthetruckwouldbemoving,thusmakingthetest morecomplicatetoadminister.Itwasdecidedtoutilizetwomethodsofintroducingpedestrian surrogatesbehindthebackingtruck: 24

PAGE 41

1.StationaryObject 2.SurprisePedestrianSurrogate 3.2.1StationaryObject Theobjectiveofthistestwastoidentifyifthesystemcouldhelpthedriversidentifythatastationaryobjectisblockingthevehicle'spathandthattoavoidcrashtheyhavetostopthevehicle.One ofthemainissueswiththistestwastomaketheobjectsmallenoughsothatisnotveryobvious, butbigenoughtohaveanimportancetothedrivertostop.Intheexperiment,anorangetrafc conewasselectedforitscontrastwiththeenvironment,andtomakeitmorevisibletothedriver.It isclearhoweverthatdifferentcolorobjectsmightnotbeasvisibleasabrightorangetrafccone. Itwasdecidedthattheconehadtobepositionedbehindthetruckafterthedriverenteredthetruck toperformthemaneuvers.Thiswouldmaketheobjectanunexpectedevent,thusprovidingamore realisticscenario.Ifthedriverwasexpectingtheobject,itwouldobviouslynotbenetthestudy. Inthisfashion,thedriverwasletinthetruckformaneuverbrieng,andduringtherstmaneuver: StraightLineBacking,anexaminerholdingastackoftwocones,walkedinaperpendicularpath behindthebackingtruckanddroppedoneoftheconesinthemiddleofthepath,asdescribedin detailinsubsection4.2.2.1.Thiswasperformedonlyforthismaneuver,sinceanactualperson carriedthecones,andthetruckwouldonlyhavetotravelatastraightline,makingiteasierto predictwherethetruckwouldbeatalltimes. 3.2.2SurprisePedestrianSurrogate Thismethodwasusedinpreviousstudiesconductingcontrolledexperiments.Thedifferencehoweverwasthattheapparatusinpreviousstudiesmovedonlyatastraightline,becausethepedestriandummyusedwashangingfromasmallcranebeingpulledbyaweight/pulleysystem.The otherdifferencewasthatthevehiclesinpreviousstudieswerestationary.Inthisstudyitwasdecidedthatthepedestriansurrogatewouldhavetobemobileandindependentofthetruck,inorder 25

PAGE 42

topositionitatrequiredtimeintervalsandpositionsinthetruck'spath.Toachievethis,athree yearoldchild'sgure3fthighwasmountedonaremotecontrolledtoy-car.Thegurewascut outfromwhitefoamboardasshowninFigure3.1andwasmuchsmallerinrelationtothetruckas showninFigure3.2. Figure3.1:Pedestriandummybehindtruck. Figure3.2:Pedestriandummyclose-up. 3.2.3TestingProcedure 1.Theparticipatingdriverswereaskedtoreadandsignadriverinformedconsentformthat explainedtheprocedureofthetest,risksanddiscomforts,benetstothedriver,extentof anonymityandcondentiality,compensation,andrights,andthustheygavetheirpermission toparticipateinthetest.TheinformedconsentformisshowninAppendixB.Thedrivers wereassignedanumberthatwouldserveasanidentierforallstagesandformsofthetest foranonymitypurposes. 2.Thedriverswerethengivenayershowingdiagramsofthethreemaneuversthatwere includedinthetest,withdetailsforeachmaneuver.Eachdriverenteredtheequippedtruck, andperformedthemaneuversinthepredenedrandomsequence.Asecondteammember wasinsidethetrucktoguidethedriverthroughthemaneuvers. 26

PAGE 43

3.Thedriverswereaskedtoperformtwosetswithandwithoutarearviewsystemofthree differentmaneuversStraightLineBacking,OffsetRightBacking,andAlleyDockBacking.Tominimizepotentialbias,theorderofthesixmaneuversforeachdriverfolloweda pre-generatedrandomnumbertable.Therefore,eachdrivercompletedthemaneuversina differentorder. 4.Anothertwoteammemberswerestationedoutsidethetrucktoobserve,record,andcontrol thetestdummies. 5.Aftercompletionofallsixmaneuvers,thedriverswereaskedtocompleteasurveytoprovidetheirfeedbackforthesystem.Theevaluationformanddatacollectionsurveyisshown inAppendixC. 3.2.4DataReduction Thedatacollectedintheexperimentwereintheformofsurveyanswersandvideofromthedrivers' behavioratthetimeofthetest.Thereductionofthesecond,thevideodata,requiredmanyhours ofwatchingvideo.Acomputersoftwarewasusedthatmultiplexesallthreecamerasandsensor informationstoredontheDataAcquisitionSystemunitandpresentsitonasinglescreen.Aview ofthisscreenisshowninFigure3.3.Thevideowasrecordedwitharateof30frames/second. Thevideowaswatchedandforeveryframeequivalentto0.033ofasecond,thedrivers'glance locationanddurationwasrecorded. Itwasdecidedthatforthedriverbehaviordata,theobjectivewastoobserveifdistinctivepatterns inthesystem'suseappearedduringtheexperiment.Thedriverswereexpectedtousethesystem likeamirror,whereduringthebackingmaneuver,thedriverscannedfromthedrivertothepassengersidemirrorwhiletryingtoadjustthetruckforitsnaldestination,avoidingpotentialhazards. Thismeantthatthelocationoftheglancesofthedriversandthedurationoftheseglancesshould berecorded.Fourdistinctlocationswereidentiedandcategorized: 1.Thedriversidemirrorandovertheleftshoulder, 27

PAGE 44

Figure3.3:Video/Sensoranalysissoftwarescreen. 2.Thepassengersidemirror, 3.Therearviewvideosystemmonitor, 4.Anyotherlocationincludingforward,leftandrightbumperconvexmirrors. ThedriverandpassengersidemirrorsincludedboththeatsurfaceWest-Coastmirrorandconvexmirrortogetherasonelocationbecauseitwasnotpossibletodistinguishfromthevideoifthe driverwaslookingatoneortheother.ThefourareasareshowninFigure3.4. 3.3StatisticalModelDevelopment Aftertheanalysisofbothbehaviorandacceptancedata,astatisticalmodelwasttedtoestimate theprobabilityofhavingabackingcrashgiventhecharacteristicscontrolledintheexperiment. Thismodelallowedinferenceontheimportantfactorsmeasuredduringtheexperimentthatinuenceif,how,andwhenabackingcrashwilloccur.Atthesametime,thismodelwasusedto 28

PAGE 45

Figure3.4:Driverfourglancelocationsusedfordataanalysis. identifythelevelswhichmakeasignicantdifferenceonifacrashwilloccurornot.Sensitivity analysisprovidedchangesinthefactors'componentsandprovidedthecorrespondingprobability change. Theoutcomeoftheexperimentwasadichotomousvariableofbackingcrashoccurrence.This wascodedas1forbackingcrash,and0fornobackingcrash.Sincethisdependentorresponse variablehadtwopossiblevalues,abinaryregressionmodelwasusedtoestimatetheprobability ofhavingacrash.Themodeleffectivelyestimatestheoutcomeasanumberbetween0and1. Withadecisionruleofcutoffvalue0.5,themodelthendecidesinwhichgrouptoplaceeachrun. 29

PAGE 46

Thiscutoffvaluecanbechangedtoreducetheerrorofmakingafalsepositiveorfalsenegative estimation. 30

PAGE 47

Chapter4:DataCollectionMethodology Thischapterdescribesthemethodologyfortherequireddatacollection.Twoareasofdatawere dened:crashdataandexperimentdata.SincethebackingcrashdatawithRVSarenotenough,in ordertoidentifythepotentialbackingcrasheswheretheRVScanbeapplied,crashrecordsneed tobereviewedtoestimatetheproblemsize,anddescription.Furthermorethedesignedexperiment providesdatapertainingtothebackingcrashinrelationtospeciccharacteristicsoftheenvironmentandthedrivers. 4.1CrashDatafromCARSystem TheFloridaCrashAnalysisReportingCARdatabaseforyears2003to2006,wasavailable forreview.Fromthedatabase,variablessuchasthevehicletypes,crashdate,time,locationand injuryseveritywereavailable.Thesevariableshoweverdonotprovideacleardescriptionofthe conditionsunderwhichthecrashoccurred.Theonlymethodtoestablishwhathappenedwas toreviewthepolicereportslongforminordertoobtaininformationonthevehiclespeedsat thetimeofcrash,andmoreimportantlythenarrativeexplanationgivenbythepoliceofceron whathappened.Thisnarrativeprovidesdetailsontheconditionsunderwhichthecrashoccurred andalsoprovidesanopportunitytocaptureanyerrorsmadeinthedatareductionfromthepolice formstobeginwith. Thetrafccrashreportformlongformisafourpageformthatthepoliceofcerslleverytime theyrespondtoacrash.Theformhastwopagesofcodedvariablesincludingvehicledata,time, location,crashnumber,vehicletypes,driverinformation,insurance,contributingcauses,vehicle 31

PAGE 48

movements,pointofimpact,rstharmfulevent,injuries,etc.Thethirdpageincludesanarrative oftheinvestigationandthefourthpageincludesaschematicdiagramofthecrash. 4.2ExperimentData Theseconddatacollectionwascompletedduringthedriverexperiment.Thisexperimentdescribedindetailinsection3.2wasdesignedtoproducebehaviordataduringbackingmaneuvers fromtruckdriverswiththeuseofRVS.Thecollectionmethodologyfortheexperimentdatais describedinthefollowingsubsections. 4.2.1DataAcquisitionSystem DuringtheexperimentthetruckwasequippedwithaDataAcquisitionSystemDAStorecord andstorethenecessarydatafortheanalysis.Thedatacollectedfortheexperimentcameintwo forms:video,audioandsensordatarecordedinthedigitalvideorecordingDVRunitand fromobservationsandmeasurementfromtheeldpersonneltakenduringthetests.Figures4.1 and4.2showtheDASunitusedintheexperiment.TheDAScomprisedofaDVR,amonitor,a GlobalPositioningSystemGPSAntenna,acoolingfan,powerswitchesandrelevantpowerand video/audioconnectors.AschematicdiagramofthecasewiringisshowninFigure4.3. Thetruckwasequippedwiththreecameras.AsshowninFigure4.4,onecameraCamera1was installedinsidethecabinnexttothemonitortorecordthedriver'sbehavior,andtworearview camerasweremountedatthebackofthetruck.ThesecondcameraCamera2wasinstalledon thetopofthecargoboxtrailerinthecenter,aimeddownatanangleof60 tothehorizontal.The thirdcameraCamera3wasinstalledundertheposition/brakelightsofthetraileratanangle of0 tothehorizontal.Camera3wasusedtoprovidelivevideofeedtothedriverthroughthe monitorlocatedinthedriver'scabin.Theblockdiagramfortheprocessingandstorageofthedata isshowninFigure4.5. 32

PAGE 49

Figure4.1:DAS-Outsideview. Figure4.2:DAS-Insideview. Figure4.3:TheDAScasewiringdiagram. 4.2.2BackingManeuvers Theexperimentwasdesignedtotesttheeffectivenessoftherearviewvideosystem,asbacking crashcountermeasure.Sincethesystemwastobetestedwhilebacking,onlycertainbacking 33

PAGE 50

Figure4.4:Locationofthethreecamerasusedintheexperiment. maneuverswereselectedtobeincludedinthenalexperiment.Thefollowingmaneuverswere 34

PAGE 51

Figure4.5:Theexperimentdataprocessing/storageprocess. includedintheexperimentafteraninitialpre-test.AllthemaneuversweretakenfromtheCommercialDriverLicenseCDLskillstest.Themaneuversare: 1.StraightLineBackingManeuver 2.OffsetRightBackingManeuver 3.AlleyDockBackingManeuver 4.2.2.1StraightLineBackingManeuver Thismaneuverrequiredthedrivertobackthetruckinastraightlineforatotalof195ft.The driverwouldhavetobackthetruckatastraightlinehencethenameofthemaneuverthrough thecones.TheschematicdiagramisshowninFigure4.6.Whenthebackofthetrailerreached about65ftfromtheendpositionthelineconnectingthersttwocones,asshown,theexaminer crossedbehindthetraileralongthepathshown.Theexaminerwhowascarryingtwo3fthigh 35

PAGE 52

cones,passedatnormalwalkingspeed,thendroppedthebottomconeinthecenterlineofthebox inthepathofthebackingtruckandcontinuedwalkingoutoftheboxwhilecarryingthesecond cone.Thisprocedurewasadoptedsothatifthedriversawtheexaminerpassingthroughtheside mirrors,hewouldnotbeabletodetectimmediatelythattheexaminerdroppedaconesinceit wouldappearthattheconewastransferredallthewaytotheothersideofthebox.Asonewould understand,ifthedriverdetectedthataconewasmissingfromthehandsoftheexaminer,the drivercouldeasilyassumethattheconewasdroppedbehindthetruckandstopforthatreason. Sincethiswasnottheintentionofthetest,theproceduredescribedabovewasused.Themaneuver wasperformedtwice,oncewithouttheuseoftheRVS,andoncewithit,inarandomorderfor eachdriver. Ifthedriverstoppedforthecone,thedistancefromthebackofthetrailertotheconewasmeasuredandrecorded.Thishappenedonbothoccasionswhenthedriverwasusingthesystem,and whenthedriverwasnotusingthesystem.Itwasexpectedthatthedriverwouldnotstopwhen thesystemwasnotbeingused.Thedriverswerenotawarewhenandiftheteammemberwould crossbehindthetrucktocreateanaturalisticreactionbythedriver.Theprocedurehadasfollows showninFigure4.7: 1.Theexaminercarriesastackoftwocones. 2.Theexaminerwalksbehindthebackingtruck,droppingaconeandcontinuestowalkaway whilestillcarryingtheothercone.Thiswasdonetopositiontheunexpectedobjectonthe backingpathofvehicle. 3.Aconeislocatedatthebackingpathofthetruck. 4.Thedriverapparentlydoesnotnoticethecone. 5.Thedrivercontinuestomakeastraightbackingmaneuverasinstructed. 6.Ifthedriverfailedtodetectthepresenceofthecone,thevehiclewouldknockitdown, signifyingacrash. 36

PAGE 53

Figure4.6:DiagramofStraightLineBackingManeuver. 37

PAGE 54

Figure4.7:StraightLineBackingManeuverconesequence. 4.2.2.2OffsetRightBackingManeuver Thismaneuverrequiredthedrivertostartfromapositionontheleftoftherstmaneuver,backing thetruckandcurvingtotherightatabox,asshowninFigure4.8.Theendoftheboxwasaloadingdock,andthedriverswereaskedtostopthetruckatadistanceof3ftfromthedoor.During thismaneuverandwhenthebackofthetrailerwaslocatedapproximately55ftfromtheloading dock,anexaminerintroducedthepedestriansurrogatedescribedinsection3.2.2.Thedummy wassteeredfromasafedistanceanditwasintroducedatapproximately20ftfromthebackof thetrailer.Ifthedriverstopped,thedistancebetweenthedummyandthetrailerwasmeasuredand recordedandtheoutcomewasnotedasastop.Ifthedriverdidnotstopandhitthedummy,then 38

PAGE 55

theoutcomewasrecordedasahit.Afterthis,thedrivercontinuedthemaneuverandthedistance stoppedfromtheloadingdockwasrecorded.ApictureofthismaneuverisshowninFigure4.9. Figure4.8:DiagramofOffsetRightBackingManeuver. 39

PAGE 56

Figure4.9:Pedestriandummyintroduction.Thedummyissteeredinthetruck'spath. 4.2.2.3AlleyDockBackingManeuver Thismaneuverrequiredthedriverstopositionthetruckata90 anglefromaboxleadingtoa loadingdockandsteerthetruckintothebox,parkingthetruck3ftfromtheloadingdock.A schematicdiagramisshowninFigure4.10.AswiththeOffsetRightBackingManeuver,when thetrailerreachedapproximately55ftfromthedock,thepedestriansurrogatewasintroducedto observethereactionofthedriver.Ifthedriverstopped,thedistancebetweenthedummyandthe trailerwasmeasuredandrecordedandtheoutcomewasnotedasastop.Ifthedriverdidnot stopandhitthedummy,thentheoutcomewasrecordedasahit.Afterthis,thedrivercontinued themaneuverandthedistancestoppedfromtheloadingdockwasrecorded. 40

PAGE 57

Figure4.10:DiagramofAlleyDockBackingManeuver. 4.2.3DeterminationofExperimentVariables Asmentionedearlier,theDASwasusedtorecordandstoredataduringtheexperimenttobe retrievedandanalyzedatalatertime.Anumberofvariableswerecreatedfortheanalysis.The variablescreatedareshowninTable4.1. Table4.1:Variablescreatedfordataanalysis. VariableSourceTypeValue OutcomeExperimentNominal1=Crash,0=NoCrash Dist_objExperimentContinuousNumberofFeet Dist_dockExperimentContinuousNumberofFeet 41

PAGE 58

Table4.1:continued VariableSourceTypeValue DriverAgeSurveyFormOrdinal1=Under25,2=26-30,3=31-40 4=41-50,5=50+ CDL_expSurveyFormContinuousNumberofYears RVS_expSurveyFormNominal1=Yes,0=No Backing_histSurveyFormNominal1=Yes,0=No Backing_noSurveyFormContinuousNumberofcrashesinpreviousyears Comf_RVS_M1SurveyFormOrdinal1=StronglyAgree,2=SomewhatAgree 3=Agree,4=Disagree, 5=SomewhatDisagree,6=StronglyDisagree Comf_help_M1SurveyFormOrdinal1=StronglyAgree,2=SomewhatAgree 3=Agree,4=Disagree, 5=SomewhatDisagree,6=StronglyDisagree Comf_RVS_M2SurveyFormOrdinal1=StronglyAgree,2=SomewhatAgree 3=Agree,4=Disagree, 5=SomewhatDisagree,6=StronglyDisagree Comf_help_M2SurveyFormOrdinal1=StronglyAgree,2=SomewhatAgree 3=Agree,4=Disagree, 5=SomewhatDisagree,6=StronglyDisagree Comf_RVS_M3SurveyFormOrdinal1=StronglyAgree,2=SomewhatAgree 3=Agree,4=Disagree, 5=SomewhatDisagree,6=StronglyDisagree Comf_help_M3SurveyFormOrdinal1=StronglyAgree,2=SomewhatAgree 3=Agree,4=Disagree, 5=SomewhatDisagree,6=StronglyDisagree Like_have_RVSSurveyFormOrdinal1=StronglyAgree,2=SomewhatAgree 42

PAGE 59

Table4.1:continued VariableSourceTypeValue 3=Agree,4=Disagree, 5=SomewhatDisagree,6=StronglyDisagree Comp_paysSurveyFormOrdinal1=StronglyAgree,2=SomewhatAgree 3=Agree,4=Disagree, 5=SomewhatDisagree,6=StronglyDisagree RVS_help_blindSurveyFormOrdinal1=StronglyAgree,2=SomewhatAgree 3=Agree,4=Disagree, 5=SomewhatDisagree,6=StronglyDisagree Man_durVideoContinuousNumberofFrames React_gl_locVideoNominalm=monitor,p=passengermirror d=drivermirror,o=otherlocation m_cntVideoContinuousNumberofGlances m_durVideoContinuousNumberofFrames p_cntVideoContinuousNumberofGlances p_durVideoContinuousNumberofFrames d_cntVideoContinuousNumberofGlances d_durVideoContinuousNumberofFrames o_cntVideoContinuousNumberofGlances o_durVideoContinuousNumberofFrames avg_transVideoContinuousNumberofFrames avg_gl_durVideoContinuousNumberofFrames 43

PAGE 60

Chapter5:CrashDatabaseAnalysis Inthisstudy,theFloridaCrashAnalysisReportingDatabaseCARyears2003-2006wasanalyzedwiththepurposeofunderstandingthecontributingfactorsanddistributionofthebacking crashesinvolvinglargetrucksand1,549actualpolicecrashreportswerereviewedtoassessthe potentialofreducingtruckbackingcrasheswiththeRVS.Alargetruckisdenedasatruckwitha GrossVehicleWeightRatingGVWRgreaterthan10,000poundsandincludesbothmediumand heavytrucks. 5.1BackingCrashes TheFloridaCrashAnalysisReportingDatabaseCARmaintainedbytheFloridaDepartment ofTransportationFDOTSafetyOfcewasusedtoretrievethecrashdatafor2003-2006tounderstandthecontributingfactorsanddistributionofthebackingcrashesinvolvingtrucks.First, theCARdatabasewasreviewedforrelevantvariablesforbackingcrash.Threevariableswere identiedthatcanbeusedtoestimatetruckbackingcrashes:thevariablecodedastherst harmfulevent:backedintowhichisusedasthecrashtype,thevehiclemovementcodedas backing,andthecontributingcauseforthecrashcodedasimproperbacking.Itisexpected thatacrashcodedasabackingcrash,withrstharmfuleventbackedintowillhaveabacking movementandthecontributingcauseasimproperbacking.However,itwasfoundthat5,718 crasheswerecodedasbackedinto,but9,092werecodedasbackingmovementand7,567 werecodedwithimproperbackingasthecontributingcause. Thisdiscrepancypromptedamorethoroughinvestigationofthedatabase,anditwasfoundthat thepoliceofcerswereresponsibleforcodingthesevariables,thusintroducinganerrorinthe 44

PAGE 61

data.Somebackingcrasheswerealsocodedtohavecarelessdriving,failuretouseduecare,unsafebacking,violationofrightofway,etcastheircause.AfterthereviewoftheCARdatabase, itwasfoundthatitisappropriatetousebothbackingastheat-faultvehiclemovementand improperbackingasthecontributingcauseofthecrashtoselectbackingcrashes. AsshowninTable5.1,backingcrashratesinFloridahaveremainedrelativelyconstantoverthe years.Itisnotedthattruckshaveahigherpercentageofbackingcrashesascomparedtopassenger cars.Therewereatotalof7,356backingcrashesthatinvolvedtrucksforfouryears. Table5.1:Percentofbackingcrashestototalcrashes,Florida-2006. Year2003200420052006Total PassengerCar Total248,215265,124270,436261,7671,045,542 Backing10,29311,16211,16811,21743,840 Percent4.154.214.134.294.19 Truck Total15,97518,46020,28319,21373,931 Backing1,6011,8001,9941,9617,356 Percent10.029.759.8310.219.95 AllVehicleTypes Total301,095323,599332,279323,1571,208,130 Backing13,70914,93315,18915,23859,069 Percent4.554.614.574.724.61 Afterreviewitseemsthatimproperbackingisamajorcontributingfactorfortruckbacking crashes.AccordingtoFloridaStatutes316.1985,adriverisprohibitedfromdoingabacking maneuverunlesssuchmovementcanbemadewithsafetyandwithoutinterferingwithothertrafc.Ifthisisviolated,thedrivercanbecitedforimproperbacking. ConsideringthatthepurposeofthisstudyistounderstandtruckbackingcrashesandpotentialreductionwiththeRVS,itisneededtoinvestigatethedetailsofbackingcrashesincludingthecrash speedandthecauseofimproperbacking.However,nofurtherinformationcanbedrawnabout thecauseofimproperbackingfromCAR.Specically,therewasnoinformationregardingif 45

PAGE 62

improperbackingcrashesoccurredduetothelackofaclearrearviewofthevehicleorwhether theimproperbackingcouldhavebeenavoidedbyprovidingarearview. Tondimproperbackingundercommoncircumstancesandwhetheritcouldbeeliminatedby providingarearviewofthevehicle,actualpolicecrashreportslongformswerereviewed.A totalof1,549individualcrashreportswerestratiedbycrashseverityandrandomlyselected. Theconditionsunderwhichthebackingcrashesoccurredweresummarizedintothefollowing cases/scenarios: 1.Thetruckdrivermissesaturn,stopsintheroad,andbackstobeabletomaketheturn,strikingavehiclethatmanagedtostopbehindthetruckwithoutthedriverseeingthevehicle. Thesecondvehicleisusuallyeitherstoppedorcomingtoastop.Noteherethatmanytruck driverssaythattheydidchecktheirmirrorspriortobackingandcouldnotseethevehicle behindthetruckduetoablindspot.Also,tractor-trailertrucks,whenbacking,giveno warningbecausethetrailersusuallyhavenobackuplightsoralarms.Straightmediumor heavytrucksusuallyhaveanaudiblealarmwhenbacking.Thiscrashoccursonroadways, intersectionswithsidestreets,drivewayaccess,andhighwayentranceorexitramps,etc. 2.Thetruckdriverrealizesisinthewronglaneatasignalandbackstochangelanes,striking thevehiclestoppedbehindit. 3.Thetruckdriverisnegotiatingaturnatanintersection,butduetothelengthofthetrailer andnothavingadequatespace,thedriverneedstobackuptocompleteturnandstrikesthe vehiclethatfollowstoocloselybehindthetruckwithinablindspot. 4.Thetruckdriverisstoppedatasignalorstopsignintersection,andtomakeroomforanotherlargevehicleturningperpendicularlywiththetruck'sdirection,backsintothevehicle behindit. 5.Thetruckdriverstopstoofar,passingthestopbaratatrafcsignalorstopsign,andbacks tocorrectthismistake,strikingthevehiclestoppedbehindthetruck. 46

PAGE 63

6.Thetruckdriverisbackinginaparkinglottodelivergoods,strikingparkedvehiclesaround itusuallybehind. 7.Thetruckdriverisbackingatanylocationandstrikesasecondvehiclebackingaswell. 8.Thetruckdriverisbackingoutofadrivewayorsidestreetintothemainstreet,strikingan oncomingvehiclethatdidnotseethetruckuntilitwastoolateorthedrivercouldnotstop intime. 9.Thetruckdriverisperformingabackingmaneuver,andthedrivercannotseearoundthe vehicle,thusstrikingavehiclethatisparkednexttoit.Thishappenswhenthedriveris turningthetractortoadjustthetrailer. Thereviewofpolicecrashreportsprovidedamixedlesson.Itwasfoundthatmorethan50percentofcrashreportsactuallyhaveadriver'sstatementregardingnoview/missedview.However itwasfoundthatthepointofimpactforbackingcrashisnotalwaystherearendoftheat-fault vehicleasitcanbecommonlyexpected.AsshowninFigure5.1,itwasfoundthatthepointof impactforat-faultvehicleisthefrontendofvehicleinmorethan10percentoftruckbacking crash. Figure5.1:Pointofimpactofat-faultvehicleinbackingcrashes. Table5.2showsthesitelocationoftruckbackingcrashes.Morethan35percentofbackingcrashes occurredinparkinglots.However,itisnotedthat14percentofbackingcrashesarealsoreported atintersectionwhereusuallybackingmaneuverisnotexpected.Also,itwasfoundthatmany crashesoccurinlocationsthattheyshouldnotbeoccurring,i.e.backingcrashesinaroadsection oratintersections.ThedistributionoflocationofthecrashesisshowninTable5.2.Asshown, 47

PAGE 64

26.4percentofbackingcrashesoccurredataroadsection,andanother14percentoccurredatan intersection.Italsoimpliesthatnotallbackingcrashesarelow-speedcrashes. Table5.2:Sitelocationofbackingcrashes,Florida-2006. MediumHeavyTruckTruck TruckormoreTractor SiteLocationreartiresrearaxlesCabTotal% RoadSection6477865271,96026.40 AtIntersection3613882841,03314.04 In.byIntersection1131581324035.48 DrivewayAccess1931931875737.79 ParkingLotPublic5203314121,26317.17 ParkingLotPrivate4323364191,18716.14 PrivateProperty1761851805417.35 *AllOther1381451133965.38 Total2,5802,5222,2547,356100 *AllOtherincludesRRXing,Bridge,Entrance/ExitRamps,TollBooth,PublicBusStop. Table5.3showstheinjuryseverityofbackingcrashesbytrucktype.Asshown,87-92percentof backingcrashesarepropertydamageonlyPDOcrashesandveryfewfatalityinvolvedcrashes werefound. Table5.3:Backingcrashinjuryseverity,Florida-2006. VehicleType MediumTruckHeavyTruckTruckTractor InjurySeverityreartires>2rearaxlesCabTotal NonePDO2,2452,2342,0816,560 PossibleInjury14913590374 Non-IncapacitatingInjury615330144 IncapacitatingInjury19301160 FatalWithin90Days2237 Non-TrafcFatality2215 Total2,5802,5222,2547,356 48

PAGE 65

5.2CrashReportsAnalysisResults Someseverebackingcrashesoccurmainlyduetoirrationaldrivingmaneuversinsteadoflimited rearviewornorearvisibility.Also,itisnotedthatthereislimitedpotentialofusingarearview videosystemtopreventcertaintypeofhighspeedbackingcrashes,includingcrashesthatoccurat a90 anglewhereonevehicleisenteringthepathofanothervehicleandthuscannotbeavoided becausethesituationleaveslittleornowarningtothedrivertoreact. Asmentionedearlier,atotalof1,549FloridatrafccrashpolicereportsLongFormbetween 2003and2006wererandomlyselectedandreviewedtoassessifthecauseofthebackingcrash wascloselyrelatedtothelackofvieworlimitedview.Thereviewofthesecrashreportsincluded backingcrashesinvolvingmediumandlargetrucksaswellastractor-trailers.Thebackingcrashes wereclassiedbyinjuryseverity,whichcanbeusedasameasureofcrashseverity.Ingeneral,the injuryseverityishighlycorrelatedwiththespeedofthevehiclesinvolvedinthecrash.Theinjury severityindexhasvecategories: 1.Noinjury 2.Possibleinjury 3.Non-incapacitatinginjury 4.Incapacitatinginjury 5.Fatality Thereviewprocessbeganwiththemostseverecrashesthatinvolvedincapacitatingandfatalinjuries.Aboutftypercentofthesecrashesinvolvedatleastonevehiclethatwastravelingathigher than10mph,andthecauseofcrashwasrandom,primarilyduetoirrationaldrivingmaneuvers insteadoflimitedornorearvisibility.Theeffortwasthenextendedtolowerinjuryseverity.The remainderofthebackingcrashesreviewedincluded104non-incapacitatinginjury,323possible injury,and1,035noinjurycrashes. 49

PAGE 66

5.2.1NoInjuryBackingCrashes FortheNoinjurycategory,codedas01,thereportsshowedthatatotalof1,080crashesinvolvedtrucksincludingvehicletype04=mediumtruck,type05=heavytruck,andtype06= tractortrailer.Ofthose,45wereeliminatedbecausetheat-faultvehicledidnotbackupbutrather rolledbackandstruckthevehiclebehindit. Forthepurposeofthisstudybeingtheuseoftherearviewvideosystemasbackingcrashcountermeasure,thecrasheswereanalyzedfortheirmajorcausesandusefulinformationfortheintended research.FirstthecrashlocationwasfoundasshowninTable5.4. Table5.4:CrashlocationforNoInjurybackingcrashes. LocationFrequencyPercent Accessroad10.10 Alley30.29 Checkpointgate30.29 Constructionsite20.19 Constructionzone10.10 Dirtroad20.19 Drivewayaccess797.63 Dumpsite10.10 Emergencylane10.10 Entranceramp50.48 Exitramp30.29 Intersection27326.38 Parkinglot33131.98 Privateproperty252.42 Railroad30.29 Shoulder20.19 Street28827.83 Tollbooth70.68 Unknown50.48 Total1,035100 Sincethecrashlocationhasquiteafewcategories,themostcommonhighestinpercentage categorieswerechosentoberepresentedinFigure5.2.Theparkinglotcategoryincludesboth 50

PAGE 67

publicandprivateparkinglots.TheothercategoryincludesallotherlocationsshowninTable5.4. Figure5.2:ThemostcommoncrashlocationsforNoInjurybackingcrashes. Nextanalysisofthespeedofvehicle1theat-faultvehicleandvehicle2thesecondvehiclewas performed.Thespeedofthevehicleistheestimatedspeedreportedbythepoliceofcer;itis notanobservedvaluebutprovidesthebestestimateforthevehiclespeedatthetimeofthecrash. Themeanofbothspeedswasbelow1.5mph,whichindicatesthatthevehiclesweretravelingat verylowspeeds.AsshowninTable5.5,74percentofthecrasheswithreportedspeedsoccurred whenboththeat-faultvehicleandthesecondvehicleweretravelingatspeedslessthan10mph. Thisincludesthecrasheswhenthesecondvehiclewasstopped,andthespeedwaszeromph. Duetothenatureofthesecrashes,thelocationandestimatedspeedofthevehicles,mostcould havebeenavoidedifarearviewwasprovidedandthedrivercouldseethattherewasavehicle 51

PAGE 68

Table5.5:EstimatedspeedofvehiclesforNoInjurybackingcrashes. Nextvehiclespeed Speedmph0-55-1010-1515-2020+Total At-fault veh.speed 0-5577178418619 5-1011010322127 10-1517151226 15-20111216 20+210025 Total7073017920783* *Discrepancyintotalsduetomissingdata. behindthetruckwhenbacking.Sincenodriverdeliberatelywantstohitanothervehicle,thelack ofrearvieworlimitedrearviewcanbenamedasthecauseofthecrash. Table5.6showsthatinthemajorityofbackingcrashes%theat-faultvehicleisatruck,but thesecondvehicleisanautomobile.Thissupportstheevidencethatsmallvehiclessuchasthe passengervehiclestendtodisappearintherearblindspotoflargervehiclessothatthedriver cannotseethempriortobacking.Thiswasreportedin133cases.9%wheretheat-faultdrivers reportedthattheycheckedtheirmirrorsbeforebackinganddidnotseethevehiclebehindthem. Table5.6:VehicletypesinvolvedinNoInjurybackingcrashes. At-FaultVehicleType OtherMediumTruckHeavyTruckTruckTractor VehicleTypereartires>2rearaxlesCabTotal Automobile214182164560 Van20231861 LightTruck575954170 MediumTruck11121538 HeavyTruck534544 TruckTractor525158 Total312312307931* *Discrepancyintotalsduetomissingdata. 52

PAGE 69

5.2.2PossibleInjuryBackingCrashes ForthePossibleInjurycategorycodedas02,theFloridaCrashDatabasehadatotalof334 crashesinvolvingtrucksvehiclecodes04,05,and06.Ofthose,11caseswereeliminatedbecausetheat-faultvehicledidnotbackupbutratherrolledbackandstruckthevehiclebehindit. Inthesecasestheat-faultdriverdidn'trealizethatthevehiclewasrollingbackandthuscould notavoidthecrashevenwithacountermeasureinplace.Forthepurposeofourstudy,thecrashes wereanalyzedfortheirmajorcausesandusefulinformationforthecountermeasureresearch.First thecrashlocationwasfoundasshowninTable5.7. Table5.7:CrashlocationforPossibleInjurybackingcrashes. LocationFrequencyPercent Airporttarmac10.31 Bridge20.62 Culdesac20.62 Drivewayaccess298.98 Exitramp51.55 Intersection11034.06 Loadingdock10.31 Medianopening10.31 Parkinglot5717.65 Privateproperty82.48 Street10231.58 Tollbooth41.24 Total323100 Sincethecrashlocationhasquiteafewcategories,themostcommoncategoriesarepresentedin Figure5.3.Theparkinglotcategoryincludesbothpublicandprivateparkinglots.Theother categoryincludesallotherlocationsshowninTable5.7. Next,ananalysisofthespeedofvehicle1atfaultvehicleandvehicle2thesecondvehiclewas performed.Thespeedofthevehicleistheestimatedspeedandisreportedbythepoliceofcer byestimationand/orreportedbywitnesses.Itisnotafact,ormeasuredinanywaybutitgivesa feeltothevehiclespeedatthetimeofthecrash.Themissingvalueswerenotreportedinthecrash 53

PAGE 70

Figure5.3:ThemostcommoncrashlocationsforPossibleInjurybackingcrashes. reports.Themeanofbothspeedsisbelow1.5mphwhichshowsthevehiclesweretravelingat verylowspeeds.Asexpectedtheat-faultvehicle,istravelingatlowerthan10mphandthesecond vehiclemostofthetimeisalsotravelingatlowerthan10mph.Table5.8showsthat53percentof thecrasheswithreportedspeedsoccurredwhenbothvehiclesweretravelingwithlessthan5mph. Thisincludesthecrasheswhenthesecondvehiclewasstoppedandthespeedwaszeromph. Table5.9showsthatinthemajorityofbackingcrashes%theat-faultvehicleisatruck,but thesecondvehicleisanautomobile.Thissupportstheevidencethatsmallvehiclessuchasthe passengervehiclestendtodisappearintherearblindspotoflargervehiclessothatthedriver cannotseethempriortobacking.Thiswasreportedin62cases%wheretheat-faultdriver reportedthattheycheckedtheirmirrorsbeforebackinganddidnotseethevehiclebehindthem. 54

PAGE 71

Table5.8:EstimatedspeedofvehiclesforPossibleInjurybackingcrashes. Nextvehiclespeed Speedmph0-55-1010-1515-2020+Total At-fault veh.speed 0-5140103218173 5-1047452967 10-1510020113 15-20510006 20+011013 Total2021611429262* *Discrepancyintotalsduetomissingdata. Table5.9:VehicletypesinvolvedinPossibleInjurybackingcrashes. At-FaultVehicleType OtherMediumTruckHeavyTruckTruckTractor VehicleTypereartires>2rearaxlesCabTotal Automobile878143211 Van67720 LightTruck25111147 MediumTruck2215 HeavyTruck1304 TruckTractor111113 Bus1304 Motorcycle1124 Other0202 Total12411175310* *Discrepancyintotalsduetomissingdata. 5.2.3Non-IncapacitatingInjuryBackingCrashes FortheNon-IncapacitatingInjurycategorycodedas03,theFloridaCrashDatabasehadatotal of104crashesinvolvingtrucksvehiclecodes04,05,and06.Similartothetwopreviousinjury severitycrashes,theNon-IncapacitatingInjurycrashesoccurredinthelocationsshowninTable5.10. ThemostcommonlocationcategoriesarepresentedinFigure5.4. 55

PAGE 72

Table5.10:CrashlocationforNonIncapacitatingInjurybackingcrashes. LocationFrequencyPercent Bridge10.96 Drivewayaccess1110.58 Entranceramp10.96 Exitramp32.88 Inuencedbyramp10.96 Intersection4341.35 Medianopening10.96 OnRoad2524.04 Privateparkinglot65.77 Privateproperty21.92 Publicparkinglot109.62 Total104100 Figure5.4:ThemostcommoncrashlocationsforNon-IncapacitatingInjurybackingcrashes. Next,ananalysisofthespeedofv1atfaultvehicleandv2thesecondvehiclewasperformed. Thespeedofthevehicleistheestimatedspeedandisreportedbythepoliceofcerbyestima56

PAGE 73

tionand/orreportedbywitnesses.AsshowninTable5.11,58percentofthecrashesoccurred whilethevehiclespeedwaslessthan10mph.Duetothenatureofthesecrashes,thelocationand estimatedspeedofthevehicles,mostcrashesabovecouldbeavoidedifthedrivercouldseethat therewasavehiclebehindthetruckwhenbacking. Table5.11:EstimatedspeedofvehiclesforNon-IncapacitatingInjurybackingcrashes. Nextvehiclespeed Speedmph0-55-1010-1515-2020+Total At-fault veh.speed 0-5403201358 5-107111515 10-15301037 15-20210003 20+2000274 Total545412387* *Discrepancyintotalsduetomissingdata. Table5.12showsthatinthemajorityofbackingcrashes%theat-faultvehicleisatruck,but thesecondvehicleisanautomobile.Thedifferenceinsizeismoreimportantincaseswherethe secondvehicleismuchsmallerthantheat-faultvehiclesuchasautomobilesormotorcycles. Table5.12:VehicletypesinvolvedinNon-IncapacitatingInjurybackingcrashes. At-FaultVehicleType OtherMediumTruckHeavyTruckTruckTractor VehicleTypereartires>2rearaxlesCabTotal Automobile18192057 Van55111 LightTruck44210 MediumTruck0000 HeavyTruck1001 TruckTractor0303 Bus2103 Motorcycle4329 Other2103 Total36362597* *Discrepancyintotalsduetomissingdata. 57

PAGE 74

5.2.4IncapacitatingInjuryBackingCrashes FortheIncapacitatingInjurycategorycodedas04,theFloridaCrashDatabasehadatotalof 60crashesinvolvingtrucksvehiclecodes04,05,and06.Ofthose,22caseswereeliminated becausetheat-faultvehicledidnotbackupbutratherrolledbackandstruckthevehiclebehind it.Another15crasheswereremovedbecauseofwrongcoding.Thecrashlocationwasfoundas showninFigure5.5. Figure5.5:ThelocationcategoriesforIncapacitatingInjurybackingcrashes. Next,ananalysisofthespeedofv1atfaultvehicleandv2thesecondvehiclewasperformed. AsshowninTable5.13,only38percentofthecrashesoccurredwhilethevehiclespeedwasless than10mph.Themajorityofbackingcrashesinthiscategoryoccurredwhenbothvehicleswere travelingwithmorethan10mph,evenmorethan20mph,thusthehighinjuryseverity. 58

PAGE 75

Table5.13:EstimatedspeedofvehiclesforIncapacitatingInjurybackingcrashes. Nextvehiclespeed Speedmph0-55-1010-1515-2020+Total At-fault veh.speed 0-551001117 5-10200002 10-15020002 15-20000000 20+200000 Total73001121* *Discrepancyintotalsduetomissingdata. Duetothenatureofthesecrashes,thelocationandestimatedspeedofthevehicles,mostcrashes abovecouldnotbeavoidedevenifthedrivercouldseethattherewasavehiclebehindthetruck whenbacking. Table5.14showsthatinthemajorityofbackingcrashes%theat-faultvehicleisatruck,but thesecondvehicleisanautomobile.Thedifferenceinsizeismoreimportantincaseswherethe secondvehicleismuchsmallerthantheat-faultvehiclesuchasautomobiles,andvans. Table5.14:VehicletypesinvolvedinIncapacitatingInjurybackingcrashes. At-FaultVehicleType OtherMediumTruckHeavyTruckTruckTractor VehicleTypereartires>2rearaxlesCabTotal Automobile38213 Van1001 LightTruck1203 MediumTruck1012 Other0404 Total614323* *Discrepancyintotalsduetomissingdata. 59

PAGE 76

5.2.5FatalBackingCrashes Thefatalbackingcrashescodedas05and06inthedatabase,weremuchlessfrequentthanother injuryseverities.Fromatotalof12fatalbackingcrashes,only8didnothavemissingdata.The mostcommonlocationsofthesebackingcrashes,areshowninFigure5.6.Thespeedofthetwo vehiclesinvolvedinthesecrashesisshowninTable5.15. Figure5.6:ThelocationcategoriesofFatalbackingcrashes. Duetothenatureofthesecrashes,thelocationandestimatedspeedofthevehicles,mostcrashes abovecouldnotbeavoidedevenifthedrivercouldseethattherewasavehiclebehindthetruck whenbacking.Thevehicletypesinvolvedinfatalbackingcrashesfortheyears2003-2006in FloridaareshowninTable5.16.Forallfatalcrashestheat-faultvehiclewasmuchlargerinsize thanthesecondvehicle. 60

PAGE 77

Table5.15:EstimatedspeedofvehiclesforFatalbackingcrashes. Nextvehiclespeed Speedmph0-55-1010-4545+Total At-fault veh.speed 0-520204 5-1001001 10-2000000 20+00123 Total21328* *Discrepancyintotalsduetomissingdata. Table5.16:VehicletypesinvolvedinFatalbackingcrashes. At-FaultVehicleType OtherMediumTruckHeavyTruckTruckTractor VehicleTypereartires>2rearaxlesCabTotal Automobile2102 LightTruck3003 Motorcycle0011 Other1001 Total6118* *Discrepancyintotalsduetomissingdata. 5.3IdenticationofApplicableBackingCrashes TopreventbackingcrasheswiththeRVS,thecrashneedstomeetcertainconditionssuchaspoint ofimpactandcrashspeed.Sincetherearviewcameraislocatedattherearofthevehicle,it wouldbedifculttoassumethattheRVScanhelptopreventbackingcrasheswhenthepointof impactoftheat-faultvehicleisthefrontsideofthevehicle.Basedoneachtypeofvehiclepoint ofimpact,thebackingcrashamountwasestimated.Resultsindicatedthatfortheat-faultvehicles, thepointofimpactusuallywastherearpartofthevehicle,whileforthesecondvehicleitwas thefrontpartofthevehicle.However,asshowninTable5.17,atotalof17.1percentofbacking crashesinvolvedvehicledamagetothefrontsideofthevehiclesimpossibleforthebackingvehicle. 61

PAGE 78

Table5.17:Pointofimpactofat-faultvehicleduringallbackingcrashes. PointofImpactFrequencyPercent FrontEnd96213.08 RightFrontCorner1411.92 RightFrontQuarterPanel791.07 RightRearQuarterPanel2062.80 RightRearCorner1,06414.46 RearEnd2,17729.59 LeftRearCorner7249.84 LeftRearQuarterPanel741.01 LeftFrontCorner781.06 Trailer1,42119.32 Other*2623.64 Total7,188100 *Note:Otherincludeshood,roof,trunk,undercarriage,windshield,overturn,etc. Also,theRVScanbeeffectivetopreventbackingcrasheswhenthevehiclespeedisrelatively low.Table5.18showsthespeedofthevehiclesatthetimeofthecrash.Asshown,84percentof backingcrashesoccurredwhenbothvehiclesarestoppingormovingwithlessthan10mph. Table5.18:Speedofvehiclesattimeofallbackingcrashes. Nextvehiclespeed Speedmph0-55-1010-1515-2020+Total At-fault veh.speed 0-57893515655900 5-101641610416210 10-1528280543 15-208322217 20+5220615 Total994583712841,185* *Discrepancyintotalsduetomissingdata. Basedontheanalysiswiththepointofimpactandspeedofvehiclesatthetimeofcrash,itseems that84percentoflowspeedbothvehiclestravelwithlessthan10mphbackingcrashesamong 83percentofallbackingcrasheswhichdamagestherearsideofthebackingvehiclecanbepotentiallyeliminatedorreducedbyadoptingRVSasacountermeasure. 62

PAGE 79

Chapter6:DriverBehaviorAnalysis Accordingtotheresearchmethodologydescribedinchapter3,thedatacollectedfromtheexperimentweretwofold:driverbehaviororusageofthesystem,anddriveracceptanceofthesystem. Theanalysisofthisdataprovidesimportantfactorsneededtobeincludedinthestatisticalmodel forthesystemevaluation. Atotalof45driversparticipatedintheexperiment.Theiragesrangedfromunder25yearsoldto above50yearsold.Theywereallmale.Figure6.1showsthe45drivers'agedistribution. Figure6.1:Thedriveragedistribution. 63

PAGE 80

Thedrivershadcommercialdrivingexperiencefrom1yearto34yearswithameanof10.91years asshowninFigure6.2.Theirdrivingexperiencedoesnotfollowtheiragedistributionpreciselyas expected,butasadrivergetsolderithasmoreexperience.Thisisduetosomedriversstartingto workatthisprofessionatanolderage.Outof45drivers15hadpreviousexperiencewiththeRVS whereasfor30driversitwasthersttimetheyusedtheRVS. Figure6.2:Drivers'CDLdrivingexperiencedistribution. Inaddition,15drivershadatleastonebackingcrashintheircareer,withonehavingupto4backingcrashes.Figure6.3showsthataccordingtotheirowntestimonials,30drivershadnobacking crashes. Thedriverbehaviorwasdividedintotwoaspects:drivervisualattentionanddriveracceptance. Thefollowingsectionspresentthedriver'sattentioninrelationtotheusageofthesystemandif thedriversacceptthedeviceasausefuladditiontotheirvehicles. 64

PAGE 81

Figure6.3:Thebackingcrashhistoryofthesampleinnumberofcrashes. 6.1DriverVisualAttention Thedriverswereexpectedtousethesysteminaspecicmanner,butinordertocollectrealistic data,thedriverswereobservedduringtheexperiment,andinferencescouldbemadeontheusage ofthesystem.ItwasfoundthatthedriversusedtheRVSjustlikeanothermirror.Theyglancedat themonitorwhilescanningfromthedriversidemirrortothepassengersidemirror.Asmentioned earlier,itwasdecidedthatfourlocationswereusedforthedriverglancelocationanalysis: 1.Driversidemirror 2.Passengersidemirror 3.Rearviewvideosystemmonitor 4.Allothers Thesefourlocationsweredistinguishedbylookingatthedriver'sheaddirectionduringthemaneuver.Figure6.4showsadriverlookingatthedriversidemirrorlocation,Figure6.5showsthe driverlookingatthepassengersidemirrorlocation,Figure6.6showsthedriverlookingatthe 65

PAGE 82

monitor,andFigure6.7showsthedriverlookingstraightforwardwhichisconsideredasother location. Figure6.4:Driversidelocation. Figure6.5:Passengersidelocation. Figure6.6:Monitorlocation. Figure6.7:Otherlocation. Wheninvestigatingthedriver'sglancelocations,apatternbecomesclear.Whenthedriversdidnot havethesystem,theyspenttheirbackingtimeglancingatthethreeareas:driverandpassenger mirrorandanyotherlocation.Atthesetimestherearofthevehicleiscompletelyblindtothe driver.Whentheyhavethesystem,however,theyhavetodividetheirtimeinfourlocations:the threementionedbeforeaswellasthemonitor.Infact,theaveragemaneuvertimeincreasedforthe maneuverswiththesystemasshowninTable6.1. Themaneuverdurationwasmeasuredfromthetimethetruckstartedmovingbackwards,until thetimethetruckstoppedattheendofthemaneuverposition.Duringthistimetherewerethree possiblecasesshowninFigure6.8: 66

PAGE 83

Table6.1:Averagemaneuvertimes. ManeuverTypeWithoutRVSsecWithRVSsec StraightLineBacking65.572.2 OffsetRightBacking60.366.0 AlleyDockBacking70.577.5 Total65.471.9 Case1:Thedriverstartedbacking,butneededtoadjustthetruck,andstopped,movedforward,stoppedagainandthencontinuedbackinguntiltheend.Ifthedriversawtheobject, hestoppedagain,similartocase2. Case2:Thedriverstartedbacking,andwhenthepedestriandummywasintroducedthe driverstoppedforacertaintime,untilthemeasurementsweremade,andcontinuedtonish themaneuver. Case3:Thedriverstartedbacking,andnotseeingthedummyorcone,thushavingacrash continuingtobackuntiltheendpoint.Inthiscasethedriverdidnotknowhehadacrash. Forallmaneuverdurationstobecomparable,thestoppedandforwardtimewassubtractedfrom thetotalelapsedtimeforalldrivers.Thereportedmaneuvertimesareonlybackingtimes. Figure6.8:Themaneuvertimeline. Thedrivershadtodividetheirbackingtimelookingatalllocations.Figure6.9showsanexample ofthesamedriverperformingtheStraightLineBackingManeuverwithandwithouttheRVS.As expected,thetimewiththeRVSwaslonger,sincetheadditionofthemonitorlocationdemanded 67

PAGE 84

theirattention.Differentdrivershoweverhaddifferentpatternsinthefrequencyanddurationof theglancelocations. Figure6.9:TheglancepatternofadriverperformingtheStraightLineBackingManeuver. Legend:p=passengerlocation,m=monitorlocation,d=driverlocation,o=other location. Thedrivers'outcomewascodedasacrashiftheystrucktheconeorpedestriansurrogate,andno crashiftheystoppedbeforestrikingthetwoobjects.First,differencesinthestoppingratesnot hittingconeordummyobjectweretestedwithandwithoutarearviewsystem.Table6.2shows thesummarizedrawdata.Totestthenullhypothesisisthesameastestingthedifferenceofthe probabilitiesofstoppingwithandwithoutaRVS: H 0 : P 1 + = P + 1 H 1 : P 1 + > P + 1 P 1 + = P 11 + P 12 = Pr Stop|WithCamera P + 1 = P 11 + P 21 = Pr Stop|WithoutCamera Thedatasetincludesthreetypesofmaneuverstests: 68

PAGE 85

Table6.2:Summarizedoutcomes. WithoutRVS StopCrash WithRVS Stop P 11 P 12 Crash P 21 P 22 1.StraightLineBackingManeuver 2.OffsetRightBackingManeuver 3.AlleyDockBackingManeuver Thesewereanalyzedusingonlytheoutcomeinformation,i.e.,crashorstop.Table6.3showsthe summarized2X2contingencytables. Table6.3:2X2Contingencytablesforthethreetests. StraightLineBackOffsetRightBackAlleyDockBackTotal WithoutRVSWithoutRVSWithoutRVSWithoutRVS StopCrashStopCrashStopCrashStopCrash WithRVS Stop 11213073597637 Crash 0135310616 Estimations: d = P 1 + )]TJ/F51 10.9091 Tf 9.991 0 Td [(P + 1 ; s 2 d = P 12 + P 21 )]TJ/F43 10.9091 Tf 9.991 0 Td [( P 12 )]TJ/F51 10.9091 Tf 9.991 0 Td [(P 21 2 n AMcNemartestwasadoptedformatchedpaireddataandtoreportthep-valuesfromthenull binomialprobabilitydistributionandtheasymptoticcalculations. McNemarTestStatistics = P 12 )]TJ/F51 10.9091 Tf 9.992 0 Td [(P 21 2 P 12 + P 21 X 2 df = 1 undernullprobabilitydistribution. Thevalueinparenthesesistheasymptoticcalculation,andtheratesandvariancesarecalculated fromtheaboveequations.AsanexamplethestoprateandvarianceoftheStraightLineBacking 69

PAGE 86

Maneuverarecalculated: StraightLineBackingIncreaseStopRate = P 11 + P 12 n )]TJ/F43 10.9091 Tf 11.187 7.385 Td [( P 11 + P 21 n = 32 45 )]TJ/F19 10.9091 Tf 11.187 7.385 Td [(11 45 = 0 : 4666 StraightLineBackingVariance = P 12 + P 21 )]TJ/F43 10.9091 Tf 9.991 0 Td [( P 12 )]TJ/F51 10.9091 Tf 9.991 0 Td [(P 21 2 n = 21 45 + 0 45 )]TJ/F43 10.9091 Tf 9.991 0 Td [( 21 45 )]TJ/F19 7.9701 Tf 13.179 4.298 Td [(0 45 2 45 = 0 : 00553 ThecalculationsoftheincreasedstoprateforallmaneuversareshowninTable6.4.Thepresence ofanRVSincreasedthestoprateofthedriversinStraightLineBackingManeuverby46.7percent,whichcanbeinterpretedastheincreaseofoddstoavoidpotentialbackingcrashinthemaneuver.Respectively,thestoprateisincreased4.4and17.8percentfortheOffsetRightBacking ManeuverandAlleyDockBackingManeuver. Table6.4:ResultofMcNemartest. ManeuverP-valueIncreasedStopRateVariance StraightLineBacking4.593e-0646.7%0.55% OffsetRightBacking0.56374.4%0.59% AlleyDockBacking0.011417.8%0.42% TotalBackTest2.274e-0623.0%0.20% Totestvariableshavingasignicanteffectontheincreaseofthestoprate,variablessuchasType ManeuverType,CamRVS,AgeDriverAge,YrCDLExperience,andExpRVSRVSExperiencewereevaluated.Themodelwasrenedfortheimportantvariables.Also,twointeraction termswereintroducedandthenalmathematicalmodelisshowninTable6.5. Table6.5:Coefcientsforrstmodel. CoefcientsEstimateStd.ErrorZvaluePr>|z| Intercept1.12850.34693.2530.00114** Type2-2.38120.4989-4.7731.81e-06*** Type3-2.51480.5091-4.9397.84e-07*** RVS-2.02930.4780-4.2452.18e-05*** Type2:RVS1.75050.71352.4530.01415* Type3:RVS-0.36861.1790-0.3130.75453 SignicantCodes:0'***'0.001'**'0.01'*'0.05'.'0.1''1 70

PAGE 87

Themathematicalmodelcanrepresenttheprobabilityofbackingcrashbasedonthecontrolled testdataasfollows: log Pr Y = 1 Pr Y = 0 = a 0 + a 1 X 1 + a 2 X 2 + a 3 X 3 + i 1 X 1 X 3 + i 2 X 2 X 3 Byenteringthecoefcients,theprobabilityofhavingacrashbecomes: Pr Y = 1 = e 1 : 1285 )]TJ/F19 7.9701 Tf 6.192 0 Td [(2 : 3812 X 1 )]TJ/F19 7.9701 Tf 6.192 0 Td [(2 : 5148 X 2 )]TJ/F19 7.9701 Tf 6.192 0 Td [(2 : 0293 X 3 1 + e 1 : 1285 )]TJ/F19 7.9701 Tf 6.193 0 Td [(2 : 3812 X 1 )]TJ/F19 7.9701 Tf 6.193 0 Td [(2 : 5148 X 2 )]TJ/F19 7.9701 Tf 6.193 0 Td [(2 : 0293 X 3 Where, X 1 = 8 > < > : 1 ; Man2 0 ; otherwise X 2 = 8 > < > : 1 ; Man3 0 ; otherwise X 3 = 8 > < > : 1 ; withRVS 0 ; withoutRVS Theprobabilityofhavingacrashhittingtheobjectasdescribedintheexperimentcanbeestimatedbyusingthemathematicalmodelforeachmaneuverasfollows: Maneuver1StraightLineBacking: Pr Crash|M1WithRVS = e 1 : 1285 )]TJ/F19 7.9701 Tf 6.193 0 Td [(2 : 0293 1 + e 1 : 1285 )]TJ/F19 7.9701 Tf 6.193 0 Td [(2 : 0293 = 28 : 89% Pr Crash|M1WithoutRVS = e 1 : 1285 1 + e 1 : 1285 = 75 : 56% ThedifferenceofcrashrateforM1with/withoutRVS=75.56%-28.89%=46.67% 71

PAGE 88

Maneuver2OffsetRightBacking: Pr Crash|M2WithRVS = e 1 : 1285 )]TJ/F19 7.9701 Tf 6.193 0 Td [(2 : 3812 + 1 : 7505 )]TJ/F19 7.9701 Tf 6.193 0 Td [(2 : 0293 1 + e 1 : 1285 )]TJ/F19 7.9701 Tf 6.192 0 Td [(2 : 3812 + 1 : 7505 )]TJ/F19 7.9701 Tf 6.193 0 Td [(2 : 0293 = 17 : 78% Pr Crash|M2WithoutRVS = e 1 : 1285 )]TJ/F19 7.9701 Tf 6.192 0 Td [(2 : 3812 1 + e 1 : 1285 )]TJ/F19 7.9701 Tf 6.193 0 Td [(2 : 3812 = 22 : 22% ThedifferenceofcrashrateforM2with/withoutRVS=22.22%-17.78%=4.44% Maneuver3AlleyDockBacking: Pr Crash|M3WithRVS = e 1 : 1285 )]TJ/F19 7.9701 Tf 6.193 0 Td [(2 : 5148 )]TJ/F19 7.9701 Tf 6.193 0 Td [(2 : 2093 )]TJ/F19 7.9701 Tf 6.192 0 Td [(0 : 3686 1 + e 1 : 1285 )]TJ/F19 7.9701 Tf 6.193 0 Td [(2 : 5148 )]TJ/F19 7.9701 Tf 6.192 0 Td [(2 : 2093 )]TJ/F19 7.9701 Tf 6.193 0 Td [(0 : 3686 = 2 : 22% Pr Crash|M3WithoutRVS = e 1 : 1285 )]TJ/F19 7.9701 Tf 6.192 0 Td [(2 : 5148 1 + e 1 : 1285 )]TJ/F19 7.9701 Tf 6.193 0 Td [(2 : 5148 = 20 : 00% ThedifferenceofcrashrateforM3with/withoutRVS=20.00%-2.22%=17.78% Furthertestingwasperformedtoevaluateiftheincreaseinthestopratewasrelatedtotheactual frequencyordurationofwatchingthemonitorduringthebackingmaneuversbythedrivers.Since thedurationofbackingmaneuverswasdifferentbytheindividualsinthetest,theupdatedfrequencyF*wasusedforfurtheranalysisasfollows: F k Driver i = F Driver i DurationofManeuver k Driver i t AverageDurationofManeuver k t 8 Drivers F k Driver i = FrequencyofDriver i ofmonitorglancesduringmaneuver k Thelikelihoodofhavingacrashwiththefrequencyofmonitorglancingwastestedandshownin Table6.6.TheODDSratioisdecreasedbythefactor e )]TJ/F19 7.9701 Tf 6.192 0 Td [(0 : 05694 = 0 : 94465. Table6.6:MonitorglancefrequencyANOVAtesting. CoefcientsEstimateStd.ErrorZ-ValuePr>|z| Intercept-1.189610.35382-3.3620.000773*** M_cnt-0.056940.03827-1.4880.136806 SignicantCodes:0'***'0.001'**'0.01'*'0.05'.'0.1''1 M cnt = F k Driver i 72

PAGE 89

Example: Pr H 11 =Probabilityofcrashwhendriverlooksatmonitor11timesduringbackingmaneuver. Pr H 10 =Probabilityofcrashwhendriverlooksatmonitor10timesduringbackingmaneuver. Pr H 11 1 )]TJ/F51 10.9091 Tf 9.991 0 Td [(Pr H 11 = 0 : 94465 Pr H 10 1 )]TJ/F51 10.9091 Tf 9.991 0 Td [(Pr H 10 Itappearsthatthelikelihoodofacrashismorecloselyassociatedwiththeratiooftimespent watchingthemonitorduringbackingmaneuver.Theprobabilityofacrashisreducedasthedurationoflookingatthemonitorisincreased.Similarlytothefrequency,thedurationofthemonitor glancewasupdatedtoM*foranalysisasfollows: M = MonitorGlanceDuration DurationofBackingManeuver ThelikelihoodofhavingacrashwiththemonitorglancedurationwastestedandshowninTable6.7.TheODDSratioisdecreasedbythefactor e )]TJ/F19 7.9701 Tf 6.193 0 Td [(0 : 1801 = 0 : 83519. Table6.7:MonitorglancedurationANOVAtesting. CoefcientsEstimateStd.ErrorZ-ValuePr>|z| Intercept-1.09050.3496-3.1190.00182 M_t-0.18010.1006-1.7910.07326. SignicantCodes:0'***'0.001'**'0.01'*'0.05'.'0.1''1 M t = Monitorglanceduration. Example: Pr H 11 =Probabilityofcrashwhendriverlooksatmonitorfor11seconds. Pr H 10 =Probabilityofcrashwhendriverlooksatmonitorfor10seconds. Pr H 11 1 )]TJ/F51 10.9091 Tf 9.991 0 Td [(Pr H 11 = 0 : 83519 Pr H 10 1 )]TJ/F51 10.9091 Tf 9.991 0 Td [(Pr H 10 Althoughthetwomeasuresmonitorglancefrequencyanddurationweretestedindividually,they arenotstatisticallysignicanttothetraditional0.05levelofsignicance. 73

PAGE 90

Fromasimpleanalysisitlookslikethethreemaneuversdidnotyieldtothesamereductionof crashprobabilities.Maneuver2OffsetRightBackingyieldedtotheminimumof4.44percent reduction,whereasmaneuver3AlleyDockBackinghad17.78percent.Thehighestreduction wasexperiencedinmaneuver1StraightLineBackingwitha46.67percentreductionofthe probabilityofhavingacrash.Apositiveoutcomeisthatareductionoftheprobabilityofhaving acrashorincreaseinstopratewasexperiencedinallmaneuvers. Ititimportanttonotethatthesizeoftherearblindspotandtherequiredmaneuvereffortforboth StraightLineBackingManeuverandOffsetRightBackingManeuverarenearlysameingeneral, butthistestintroducedadummyobjectwithdifferentmethods.FortheStraightLineBacking Maneuver,thedummyobjectwaspositioneddirectlybehindthevehiclewhileadummyobject wasmovedfromtheeithersideofthevehicletotherearfortheOffsetRightBackingManeuver. Whentheobjectwasapproachedfromtheside,thedriverhadachancetospotitwithoneofthe twosideviewmirrors,buttheobjectwouldberelativelydifculttodetectwhenpositionedfrom therightrearofthetruck.Therefore,thedifferenceintheresultsshouldnotbeinterpretedasa differenceinmaneuvers.Rather,itrepresentsadifferentlevelofeffectivenessoftherearview videosystemtodetectobjectsfromdifferentsidesofthevehicle. Also,itisnotedthatthereisadifferenceintheincreaseofthestoprateforOffsetRightBacking ManeuverandAlleyDockBackingManeuver,althoughthesamemethodwasusedtointroduce adummyobjectintothetest.ItappearsthattheAlleyDockBackingManeuverrequiresmore attentionandeffortbythedrivertocontrolthevehiclecomparedtotheOffsetRightBacking Maneuver,anditcanresultinmoreopportunitiesforthedrivertomisstheapproachingobject withthetraditionalsideviewmirrors. Thisanalysishoweverhasitsmainobjectivetoincludethedriverbehaviorandusageofthesystemwhilebacking.Tocontinuethedriverbehavioranalysis,thevariablesshowninTable6.8were tabulatedinacorrelationmatrixtoobtaincorrelationsbetweenthem. Sincearegressionmodelwastobeutilizedtocalculatetheprobabilityofhavingacrash,thevariablesneededtobeuncorrelated.Manyoftheabovevariableswerehighlycorrelatedsotheycould 74

PAGE 91

Table6.8:Variablesusedforbehavioranalysis. VariableSourceTypeValue OutcomeExperimentNominal1=Crash,0=NoCrash DriverAgeSurveyFormOrdinal1=Under25,2=26-30,3=31-40 4=41-50,5=50+ CDL_expSurveyFormContinuousNo.ofYears RVS_expSurveyFormNominal1=Yes,0=No Backing_histSurveyFormNominal1=Yes,0=No Backing_noSurveyFormContinuousNumberofCrashes Man_durVideoContinuousNumberofFrames m_cntVideoContinuousNumberofGlances m_durVideoContinuousNumberofFrames p_cntVideoContinuousNumberofGlances p_durVideoContinuousNumberofFrames d_cntVideoContinuousNumberofGlances d_durVideoContinuousNumberofFrames o_cntVideoContinuousNumberofGlances o_durVideoContinuousNumberofFrames avg_transVideoContinuousNumberofFrames avg_gl_durVideoContinuousNumberofFrames notbeusedinaregressionmodel.Aprincipalcomponentanalysiswasthenutilizedtosolvethis problem.Accordingto[30]: PrincipalComponentAnalysisPCAinvolvesamathematicalprocedurethattransformsasetofcorrelatedvariablesintoasmallersetofuncorrelatedvariablescalled principalcomponents. Inaddition,thevariablesrelatedtothedrivers'glancescanbegroupedintoacategorydescribing thedriver'svisualattention.ThestatisticalpackageSPSS R v17.0wasusedtoperformtheprincipalcomponentanalysisandregression. Kaiser'sMeasureofAdequacyMSAforavariable X i istheratioofthesumofthesquaredsimpler'sbetween X i andeachotherX.Thus, MSA = S r 2 ij S r 2 ij + S pr 2 ij TheoverallMSAforthesevariablesis0.567whichisadequate. 75

PAGE 92

Table6.9:Correlationmatrixforthe8driverattentionvariables. Variablesm_cntm_td_cntd_tp_cntp_to_cnto_t m_cnt1.0000.9070.437-0.1950.135-0.201-0.020-0.056 m_t 0.9071.0000.331-0.1700.023-0.241-0.015-0.052 d_cnt 0.437 0.3311.000-0.0710.80070.3070.3780.253 d_t-0.195-0.170-0.0711.000-0.235-0.2330.0880.061 p_cnt0.1350.023 0.800 -0.2351.0000.6280.2670.186 p_t -0.201 -0.241 0.307 -0.233 0.6281.0000.1690.190 o_cnt-0.020-0.015 0.3780.088 0.2670.1691.0000.780 o_t-0.056-0.052 0.2530.0610.1860.190 0.7801.000 *Note:Shadedcellsrepresenthighcorrelations. Nexttheprincipalcomponentswereextracted.Theprincipalcomponentsareuncorrelatedcomponentsthatexplainthevarianceofthecorrelatedvariables.Table6.10showstheeigenvaluesand proportionsofvariancefortheeightcomponents. Table6.10:Totalvarianceexplained. InitialEigenvalues ComponentTotal%ofVarianceCumulative% 12.69733.71533.715 22.15226.90060.615 31.48418.55479.169 40.87610.95190.120 50.4155.18595.305 60.2052.56097.865 70.0891.11098.975 80.0821.025100.000 ExtractionMethod:PrincipalComponentAnalysis. Inordertodecidehowmanycomponentstoretain,aruleofthumbistoretainonlycomponents witheigenvaluesofoneormore.Thatis,dropanycomponentthataccountsforlessvariance thandoesasinglevariable.Atesthowevertohelpwiththisdecision,theSCREEtestisaplot thatvisuallyaidsdecidingatwhatpointincludingadditionalcomponentsnolongerincreases 76

PAGE 93

theamountofvarianceaccountedforbyanontrivialamount.Figure6.10showstheSCREEplot providedbySPSS. Figure6.10:TheSCREEplotprovidedbySPSS. Forthesedata,itwasdecidedthatonlytherstthreecomponentswouldberetained.Component fourcouldbeincluded,butsincetherewereonlyeightvariables,andwiththreecomponentsexplainingalmost80percentofthevariance,itwasdecidedtobedropped.AfteraVarimaxrotation, therotatedcomponentmatrixisshowninTable6.11.Thecoefcientscanbeusedtocreatethe threechosencomponents. Thethreecomponentswerenamedbasedontheirmagnitude: Component1:monitorlocationcomponent Component2:driverandpassengerlocationcomponent Component3:otherlocationcomponent 77

PAGE 94

Table6.11:Rotatedcomponentmatrix.* Component Variable123 Monitorglancefrequency0.9660.079-0.039 Monitorglanceduration0.954-0.024-0.037 Passengerglancefrequency0.0930.9090.208 Passengerglanceduration-0.3320.8050.086 Driverglancefrequency0.4550.6650.375 Driverglanceduration-0.172-0.4870.332 Otherglancefrequency0.0160.1290.918 Otherglanceduration-0.0470.0780.895 ExtractionMethod:PrincipalComponentAnalysis. RotationMethod:VarimaxwithKaiserNormalization. *Note:Rotationconvergedin4iterations. Asmentionedearlierthethreenewcomponentsaccountfor79.17percentofthevarianceofthe eightinitialvariables. Toensurethatthethreecomponentsarenotcorrelated,thecorrelationmatrixwascalculatedand showninTable6.12.Thehighcorrelationsexperiencedbeforearesignicantlyreducedafterthe PrincipalComponentAnalysisandextractionoftherstthreeprincipalcomponents. Table6.12:Principalcomponentscorrelationmatrix. Driverand MonitorPassengerOther ComponentLocationLocationLocation MonitorLocation1.0000.161-0.119 DriverandPassengerLocation0.1611.0000.162 OtherLocation-0.1190.1621.000 Abinarylogisticregressionmodelwasestimatedwiththethreenewcomponentsgroupedfor drivers'attention,thedriverage,previousexperiencewiththeRVS,backingcrashhistory,the maneuvertype,iftheRVSwasONorOFFandtwointeractiontermswithRVSandmaneuver type.Dummyvariableswerecreatedforallcategoricalvariablesandoneofthelevelswaskeptas theirbase.Fordriverage,theagelevelabove50wasthecontrol.Formaneuvertype,maneuver 78

PAGE 95

3wasthecontrolsincemaneuvers2and3hadasimilarpedestriandummyintroduction,whereas maneuver1wasdifferent.Theregressionmodelispredictingthelogit,thatis,thenaturallogof theoddsofhavingoneortheotheroutcomecrash,not-crash.Thatis, ln ODDS = ln Y 1 )]TJ/F19 10.9091 Tf 11.628 2.171 Td [( Y = a 0 + a 1 X 1 + a 2 X 2 + ::: + a i X i Where, Y isthepredictedprobabilityoftheeventwhichiscodedwith1crashratherthanwith0 no-crash, 1 )]TJ/F19 10.9091 Tf 12.043 2.171 Td [( Y isthepredictedprobabilityoftheotheroutcome, X i isthepredictorvariable with i variablesand a arethecoefcientsinthemodel.The-2LogLikelihoodstatisticofthe modelis202.670.ThevariablesintheequationoutputshowninTable6.13,showtheregression equationtobe: ln ODDS = 4 : 585 )]TJ/F19 10.9091 Tf 9.991 0 Td [(7 : 051 Age < 25 )]TJ/F19 10.9091 Tf 9.991 0 Td [(8 : 171 Age 26 )]TJ/F19 10.9091 Tf 9.991 0 Td [(30 )]TJ/F19 10.9091 Tf 9.991 0 Td [(7 : 081 Age 30 )]TJ/F19 10.9091 Tf 9.991 0 Td [(40 )]TJ/F19 10.9091 Tf 8.476 0 Td [(7 : 869 Age 40 )]TJ/F19 10.9091 Tf 9.992 0 Td [(50 )]TJ/F19 10.9091 Tf 9.991 0 Td [(0 : 787 RVS Exp Yes )]TJ/F19 10.9091 Tf 9.992 0 Td [(1 : 671 RVS ON )]TJ/F19 10.9091 Tf 8.476 0 Td [(0 : 292 Crash history + 3 : 887 Man 1 + 1 : 292 Man 2 )]TJ/F19 10.9091 Tf 8.476 0 Td [(0 : 076 PC 1 )]TJ/F19 10.9091 Tf 9.991 0 Td [(0 : 053 PC 2 + 0 : 040 PC 3 + 1 : 322 RVS ON Man 1 + 2 : 446 RVS ON Man 2 Themodel'sdecisionruleisusedtodetermineintowhichgrouptoclassifyeachrungiventhe testsestimatedprobabilityofcrash.Themostobviousdecisionrulewouldbetoclassifytherun intothecrashgroupif r > 0 : 5andintothenocrashgroupif r < 0 : 5.Sincewedonotwantto maketheerrorofnotpredictingacrashwhenitoccurred,thedecisionrulewaschangedto0.4cutoffvalue.Thisincreasedtheoverallcorrectpercentageto81.5percentandsignicantlyincreased thepercentcorrectforcrashprobabilityestimation.Table6.14showsthespecicationforthetwo models.Thesecondwithcutoffvalue0.4ispredictingcorrectlymorecrashoccurrences. WecannowusethismodeltopredicttheODDSthatadriverwithgivencharacteristicswillhave abackingcrash.TheODDSpredictionequationis ODDS = e a 0 + a 1 X 1 + a 2 X 2 + ::: + a i X i .Ifwehave adriverthatis25yearsold,withnopreviousRVSexperience,withonepreviousbackingcrash, 79

PAGE 96

Table6.13:Variablesinthemodel. VariableBS.E.WaldSig. Constant4.5852.7542.7710.096 Age1-7.0512.6736.9600.008 Age2-8.1712.8048.4940.004 Age3-7.0812.6577.1010.008 Age4-7.8692.6828.6060.003 RVSExpYes-0.7870.4203.5140.061 RVSON-1.6711.1951.9540.162 Crashhistory0.2920.2661.2050.272 Man13.8870.70330.5570.000 Man21.2920.6993.4200.064 PC1-0.0760.0277.6640.006 PC2-0.0530.01512.5570.000 PC30.0400.0291.8910.169 RVSONbyMan11.3221.3181.0060.316 RVSONbyMan22.4461.3513.2750.070 Table6.14:Classicationstatisticsfordifferentdecisionrules. CutoffValue0.50.4 Sensitivity0.4930.634 Specicity0.9200.879 FalsePositiveRate0.3140.651 FalseNegativeRate0.1640.129 Sensitivity=percentageofoccurrencespredictedcorrectly,Specicity=percentageofnon occurrencespredictedcorrectly,Falsepositiverate=percentageofpredictedoccurrenceswhich areincorrect,Falsenegativerate=percentageofpredictednonoccurrenceswhichareincorrect. withtheRVSOFFperformingmaneuver1,wewouldhavetherstequation,whereasthesame driverwithallvariablesconstantexcepttheRVSisONwewouldhavethesecondequation: ODDS = e 4 : 585 )]TJ/F19 7.9701 Tf 6.193 0 Td [(7 : 051 + 0 : 292 + 3 : 887 )]TJ/F19 7.9701 Tf 6.193 0 Td [(0 : 076 )]TJ/F19 7.9701 Tf 6.192 0 Td [(0 : 053 + 0 : 040 = 5 : 073 ODDS = e 4 : 585 )]TJ/F19 7.9701 Tf 6.193 0 Td [(7 : 051 )]TJ/F19 7.9701 Tf 6.193 0 Td [(1 : 671 + 0 : 292 + 3 : 887 )]TJ/F19 7.9701 Tf 6.193 0 Td [(0 : 076 )]TJ/F19 7.9701 Tf 6.192 0 Td [(0 : 053 + 0 : 040 + 1 : 322 = 3 : 579 80

PAGE 97

TheODDScanbeconvertedtoprobabilitiesusing: Y = ODDS 1 + ODDS Usingtheaboveequationtheprobabilityofhavingacrashforthedriverdescribedaboveduring theexperiment,withtheRVSOFFis83.53percent,whereasthesamedriverwiththeRVSON hasaprobabilityof78.16percent.Observingthesignsofthecoefcientsonecanalsoinferonthe differentperspectivesofthemodel.Asexpected,adriverwithpreviousRVSexperiencewillhave asmallerprobability,adriverwithmorebackingcrasheswillhaveanincreasedprobability,and inthisparticulartest,themaneuver1hasahigherprobabilityandmagnitudethantheothertwo maneuvers.Certainvariablesdidnotcomeassignicantasothersbutwerekeptinthemodelfor controlpurposes. ThesecondmodelshowninTable6.13ismeanttohelpwiththeidenticationofimportantfactors thatinuencetheuseofthesystem.Althoughitcalculatestheprobabilityofhavingabacking crashasdescribedintheexperiment,thismodelisnotmeanttobeusedfortheestimationofcrash probabilityingeneral.Theprobabilitiescalculatedareratherhigh.Itmustbekeptinmindthat thismodeldescribesthedatacollectedduringthisexperiment.Itisfarfromcompleteandthe probabilityestimatesdescribetheconditionsunderwhichtheparticulardriversparticipatingin theexperiment,performedtherequestedmaneuvers.Thedatacollectedduringthisexperimentare certainlylimited.AlthoughageneralestimationoftheeffectivenessoftheRVScanbeexplained, therelationshipsbetweenthefactorsincludedinthemodelarenotinvestigatedsincethedataare notsubstantial.Onecanalsonotethatsomevariablesincludedintheregressionmodelingarenot statisticallyassignicantasothers.Theapproachofincludingsensibleandquestionablevariables wasfollowedinsteadofthepureoptimizationofthebestmodelt. 81

PAGE 98

6.2DriverAcceptance Thedriveracceptanceofanyin-vehiclesystemisasimportantastheusage,becauseunlessthe deviceisacceptedbythedriver,itwillneverreachitsfullpotentialunlessitisfullyautomatic.In thecaseoftheRVSandthisstudy,itwasfoundthattruckdriverswerenotconvincedoftheusefulnessofthesystemfromthebeginningbutratherwantedtotestthesystemrst.Asmentioned earlier,afterthetestwascompleted,aquestionnairewasgiventotheparticipatingdriversshown inAppendixC.Thedriverswereaskedaseriesofquestionsabouthowtheyfeltbackingwiththe systeminrelationtohavingnosystem.Therepliescouldnotbeincorporatedintothestatistical modelshownintheprevioussectionsotheyarepresentedseparatelyhere. Firstthedriverswereaskedtosayiftheyfeltmorecomfortableperformingthebackingmaneuver withtheRVSthanwithout.Therepliesfromthisquestionforthethreemaneuversarepresentedin Figures6.11,6.12,and6.13respectively. Figure6.11:Graphshowingrepliestoquestion:WiththeRVSIfeelmorecomfortableperforming theStraightLineBackingManeuver.Ifthe6levelsaregroupedintotwoAgreement, Disagreementthen87%ofthedriversagreewiththestatement. 82

PAGE 99

Figure6.12:Graphshowingrepliestoquestion:WiththeRVSIfeelmorecomfortableperforming theOffsetRightBackingManeuver.Ifthe6levelsaregroupedintotwoAgreement, Disagreementthen73%ofthedriversagreewiththestatement. Inaddition,thedriverswereaskedifthesystemhelpsthemavoidpotentialhazardswhileperformingthemaneuvers.TherepliestothesequestionsforthethreemaneuversareshowninFigures6.14,6.15,and6.16. Thedriverswerealsoaskediftheywouldliketohavethissystemonthetruckfortheireveryday drivingoperations.AsshowninFigure6.17,78percentofthemrepliedYesingeneral.The driverswerealsoaskediftheRVShelpsinreducingtherearblindspotoftheirvehicle,whichis themainpurposeoftheRVS.InFigure6.1893percentofthedrivers-thehighestpercentageagreedwiththisstatementafterthecontrolledtest. 83

PAGE 100

Figure6.13:Graphshowingrepliestoquestion:WiththeRVSIfeelmorecomfortableperforming theAlleyDockBackingManeuver.Ifthe6levelsaregroupedintotwoAgreement, Disagreementthen80%ofthedriversagreewiththestatement. 84

PAGE 101

Figure6.14:Graphshowingrepliestoquestion:TheRVShelpstoreducepotentialcrashesduring theStraightLineBackingManeuver.Ifthe6levelsaregroupedintotwoAgreement, Disagreementthen91%ofthedriversagreewiththestatement. 85

PAGE 102

Figure6.15:Graphshowingrepliestoquestion:TheRVShelpstoreducepotentialcrashesduring theOffsetRightBackingManeuver.Ifthe6levelsaregroupedintotwoAgreement, Disagreementthen82%ofthedriversagreewiththestatement. 86

PAGE 103

Figure6.16:Graphshowingrepliestoquestion:TheRVShelpstoreducepotentialcrashesduring theAlleyDockBackingManeuver.Ifthe6levelsaregroupedintotwoAgreement, Disagreementthen89%ofthedriversagreewiththestatement. 87

PAGE 104

Figure6.17:Graphshowingrepliestoquestion:WouldyouliketohavetheRVSonthetruckyou driveeveryday.Ifthe6levelsaregroupedintotwoYes,Nothen78%ofthedriverssaidYes tothequestion. 88

PAGE 105

Figure6.18:Graphshowingrepliestoquestion:TheRVShelpstoreducetherearblindspotofthe vehicle.Ifthe6levelsaregroupedintotwoAgreement,Disagreementthen93%ofthedrivers agreewiththestatement. 89

PAGE 106

6.3AnalysisResults Theresultsoftheanalysisshownintheprevioussectionwillbesummarizedanddiscussedinthis section. 6.3.1DriverVisualAttentionResults ThevisualattentionanalysisshowedthatthedriversrequiredmoretimetoperformthemaneuverswiththeRVS.WhentheRVSwasOFF,thedriversdividedtheirattentionunevenlywiththe highesttimespentonthedriversideviewmirror,lessonthepassengermirrorasexpected,andthe restofthetimeontheotherbumper-convexmirrors.WhentheRVSwasON,thedriversdivided theirattentionsimilarlytothedriver,passengerandmonitorlocations,increasingslightlytheir attentiontothedriverandpassengermirrors,anddecreasingtheirtimeontheotherlocations. Thisincreaseddemandonthemonitor,accountedfortheincreaseinmaneuvertime.Figure6.19 showsthemeanglancespersecondforthetwogroups. Figure6.19:Drivermeanglancepersecondduringexperiment. 90

PAGE 107

Inaddition,thedriverglancesrequiredmoretimethananyotherglancelocation,mainlydueto thedriverslookingovertheshoulderoroutsidethedriverwindowcapturedinthislocation.The passengerlocationissecond,withthetransitiontimethetimebetweenglancessharedalmost equallywiththemonitorlocation.Itseemsthatthedriversareabletoadjusttheirglancingtime withanadditionallocation.Figure6.20showsthedistributionofthemeantimespentduring glances. Figure6.20:Drivermeantimeperglance.Thetimeismeasuredinseconds.Transitionisthetime requiredforthedrivertoturnitsheadfromonelocationtotheother,i.e.betweenglancelocations. Thestatisticalmodelpresentedintheprevioussectionshowsthattheprobabilityofhavingacrash ascapturedintheexperiment,wasdecreasedwhentheRVSwasused.Differentrangesofreductionwererecordedwiththedifferentmaneuversutilizedastests.Ingeneral,thereisaminimumof 4.44percentandamaximumof46.67percentreductionofprobabilityofabackingcrash. 6.3.2DriverAcceptanceResults Theanalysisperformedforthedeviceacceptanceofthedeviceisqualitativesincenostatistical inferencecanbedrawnfromthesample.Overallthedriversseemtoagreethattheyfeltmore 91

PAGE 108

comfortableperformingthebackingmaneuverwiththeRVSthanwithout.Thequestionswere askedforeachmaneuverseparatelytocaptureanybiastowardsagivenmaneuver.About87percentofdriversagreedthattheyfeltmorecomfortableperformingtheStraightLineBackingManeuver,whereas73percentagreedfortheOffsetRightBackingManeuverand80percentforthe AlleyDockBackingManeuver.FurthermorewhenthedriverswereaskediftheRVShelpsreduce potentialbackingcrashes,91percent,82percentand89percentagreedfortheStraightLineBacking,OffsetRightBackingandAlleyDockBackingmaneuversrespectively.Also,themajorityof drivers,about78percent,saidthattheywouldwanttohavetheRVSonthetrucktheydriveevery dayand93percentagreedthattheRVShelpsinreducingtherearblindspotofthevehicle.These percentagesshowthatthedriversoverallacceptthedevice,exceptthefewcaseswhodidnotlike thedevice,anduseditverylittleordidnotuseitwhenitwasON.Figure6.21showsthatduring theexperiment,17.8percentofdriversdidnothaveacrash,whereas35.6percentexperiencedone crash,24.4percenthadtwocrashes,17.8percenthadthreecrashes,and2.22percenthadbothfour andvecrasheseach. 6.3.3AccuracyTestResults Asmentionedinsection4.2,duringthetestandwhenthedriversstoppedattheintroductionof thepedestriansurrogate,thedistancebetweenthedummyandthetrailerwasmeasured.Alsofor accuracydetermination,formaneuvers2and3,thedriverswereaskedtostopthetruckwhenthe rearofthetrailerwaslocatedat3ftfromtheloadingdock.Themeasurementswerealsorecorded andthedistancesweretestedforsignicantdifferences.OnewouldexpectthatthedistancebetweenthedummyandthetrailerwouldbesimilarorgreaterwhentheRVSwasused,whereasthe distancefromthedockshouldbeshorterforthemaneuversperformedwiththeRVS.Apaired t-testwasperformedonthedataandtheresultsareshowninTable6.17.Table6.15showsthe descriptivestatisticsandTable6.16showsthecorrelationsbetweenthepairs. Itseemsthatonlypairs3and5experiencesomewhatahighandsignicantcorrelationbetween them.Thet-testhowevershowsthatonlyforpair2:O2dist.&W2dist.existsevidencethatthe 92

PAGE 109

Figure6.21:Backingcrashesexperiencedduringexperiment. Table6.15:Pairedsamplesdescriptivestatistics. VariablePairsMeanNStd.DeviationStd.ErrorMean Pair1:O1dist.17.5155117.742242.33437 W1dist.14.2882117.887502.37817 Pair2:O2dist.11.0360303.251210.59359 W2dist.9.3977303.118120.56929 Pair3:O2dock6.5236452.118720.31584 W2dock6.2653452.285180.34065 Pair4:O3dist.10.0591353.475710.58750 W3dist.10.3049353.807730.64362 Pair5:O3dock6.6051452.341430.34904 W3dock6.1033452.450280.36527 O1=Maneuver1withoutRVS,W1=Maneuver1withRVS,dist.=distancefromdummyto trailer,dock=distancefromdocktotrailer,etc. meansaresignicantlydifferentandcanbeattributedtotheuseoftheRVS.Forallotherpairs thereisnotstrongevidencetosuggestthatthedifferenceinmeansisattributedtotheRVS. 93

PAGE 110

Table6.16:Pairedsamplescorrelations. PairsNCorrelationSig. Pair1:O1dist.&W1dist.110.4380.178 Pair2:O2dist.&W2dist.300.1450.443 Pair3:O2dock&W2dock450.5820.000 Pair4:O3dist.&W3dist.350.3370.048 Pair5:O3dock&W3dock450.5910.000 Table6.17:Pairedsamplestest. Std.Std.Error PairsMeanDeviationMeantdfSig.-tailed Pair1:O1dist.&W1dist.3.227278.283942.497701.292100.225 Pair2:O2dist.&W2dist.1.638334.164490.760332.155290.040 Pair3:O2dock&W2dock0.258222.019200.301010.858440.396 Pair4:O3dist.&W3dist.-0.245714.202070.71028-0.346340.732 Pair5:O3dock&W3dock0.501782.169490.323411.552440.128 AsaresultoftheaccuracytestthereisevidencethattheRVShelpsdriversbackthetrucktoa shorterdistancetothetargetthanwithouttheRVSinonlyoneofthemaneuvers.However,if oneobservestheoverallaveragedistance,thedriverswiththeRVSstoppedthetruckatshorter distancesthanwithoutRVSasshowninTable6.18. Table6.18:Distancetodummyanddockformaneuvers. O1W1 O2W2 O2W2 O3W3 O3W3 dist.dist. dist.dist. dockdock dist.dist. dockdock n1132 3537 4545 3644 4545 min3.832.25 3.832.67 0.831.25 5.001.00 1.170.83 max26.9232.00 35.0037.00 45.0045.00 36.0044.00 45.0045.00 avg 15.96 15.00 11.37 10.43 7.29 7.08 10.51 10.89 7.34 6.90 94

PAGE 111

Chapter7:GeneralImplementationforCamera/Video-BasedDevices Allcamera/videoimagingsystems,eitherinfontainmentorsafetyorientedprovideinformation tothedriverinrealtimeandmostrequirethedriver'sattention.Especiallyinthecaseofsafety devices,thedriverneedstopayattentiontothedevice,andrespondaccordingly.Thisshould nothoweverhappeninexpensetothevisualormentalattentionrequiredtoperformthemain task:driving.Anydeviceshouldadheretocertainrulesandspecicationswhichfollowcommon senseandinthiscasefollowrecommendationsfromthestudydescribedinthispaper.Threekey elementswereidentiedandarebrieycovered: Scientic/EngineeringGuidelines UserAcceptance UsagefromDrivers Theseareexplainedinmoredetailbelow. 7.1Scientic/EngineeringGuidelines Asdescribedin[31],somegeneralguidelinesshouldbefollowed. Thedeviceshouldperformwhatitisintendedto,operationallye.g.,presentdriverarear viewofthevehicleandconceptuallye.g.,coverblindrearspot. Thedeviceshouldpresenttheimageinrealtime. Thedeviceshouldbeconsistentinitsrepresentationoftherearsceneovertime. 95

PAGE 112

Thedeviceshouldbeabletooperateaccuratelyandreliablyinbothdaytimeandnighttime illuminationconditions. Fordevicesthatproduceaudiblewarnings,itshouldbepossibletoheartheauditoryoutput underalldrivingconditionsatalevelthatisnotstartlingtotheuser.Thevolumeofauditory outputshouldbeadjustableoverareasonablerange,approximately50dBto90dB. Thedeviceshouldbeabletooperateaccuratelyandreliablyovertheexpectedrangeof truckcabtemperature,humidity,andvibrationconditions. Thedeviceshouldbeabletooperatecontinuallyandrobustlyovertimewithonlynormal maintenanceandreplacementcosts. Inaddition,commonspecicationrelatedtovideosystems,aredescribedindetailin[26]and include: SystemoperatingTemperatureRange,ExposureTemperatureRange ExteriorIlluminationOperatingRange CabInteriorIlluminationOperatingRange MonitorLuminance/ContrastAdjustment MonitorLuminanceFluctuation SystemMinimumImageResolution SystemRefreshRate SystemMaximumImageDelay SystemPersistence SystemReliability VibrationandShockImmunity 96

PAGE 113

ResponsetoActivation ImageAspectRatio Thereaderisreferredtotheabovepublicationfordetailsonthesespecications. 7.2UserAcceptanceAssessmentMethodology Regardlessoftheprojectedsafetybenetforanygivencamera/videoimagingdevice,successful deploymentisunlikelyifusersdonotdeemthedeviceacceptable.Pastworkandmorerecent workinthearea,hasspeciedanumberofcriteriathatshouldbeconsideredwhenevaluating theuseracceptanceofsuchtechnologies.Accordingto[32]surveysfrommotorcarriers,drivers, manufacturers,andinsurancecompaniesshowedthatimportantfactorsindecisionstomake,purchaseanduseon-boardsafetytechnologiesincludedreturnoninvestmentforpurchaser,demonstratedeffectivenesstoimprovesafety,reliabilityandmaintainability,liability,marketdemand, initialcost,marketimage,driveracceptanceandin-cabinterfacetechnologyintegration.Inmany casesforheavycommercialapplications,inowner-operatorcarriersthedriveristhesameperson decidingtopurchasethedevice.Oneofthefactorsusuallynotconsidered,isthedriveracceptance.Evenifallothersareproven,thedevicewillnotbeutilizedunlessthedriverwantsto.In thecaseofvideosystemswhichrequirethedriver'sattention,thedrivercaneasilynotusethe devicethusgainingnothingfromhavingitonthevehicle. Accordingto[31]buildingonandencompassingtheseeffortswherepractical,theDepartmentof TransportationDOThasconceptualizedamethodologytosystematicallyassessuseracceptance forthepurposesoftheevaluationofvariousnewandemergingvehicletechnologies.ThismethodologyislargelybasedontheNHTSAITSStrategicPlanfor1997-2002[33]andhasevolvedand beenexpandediterativelyforDOTprojectsinvolvingeldoperationaltests.Inthisapproach, acceptanceisdependentuponthedegreetowhichadriverperceivesthebenetsderivedfroma systemasgreaterthanthecosts.Ifasystem'spotentialforsafetyisnotperceivedtooutweigh itscosts,itislikelythatthesystemwillnotbepurchased,orpurchasedbutnotutilized.Onthe 97

PAGE 114

otherhand,ifsafetypotentialanddrivingskillenhancementasrelatedtodeviceuseareperceived, thenthereisachancethatuserswillfeelcomfortableengaginginriskierdrivingbehavior.Itis importantthateachoftheseoutcomesbeassessedforacompleteevaluationofthesafetyanduser acceptanceofsuchtechnologies.Thismaybeaccomplishedconceptuallybydeconstructinguser acceptanceintovebroadelements:easeofuse,easeoflearning,perceivedvalue,advocacy,and driverbehavior.Systematicevaluationoftheseareasincludestheassessmentofmultiplecriteria, eachofwhichisbasedonhumanfactorsprinciplesasappliedtovehicletechnologies. Easeofuseisonecomponentofuseracceptanceforavehiclesafetydevice.Thisencompasses thedegreetowhichdriversndatechnologyunderstandable,usable,andintuitiveinitsoperation andmaintenance.Fullconsiderationshouldbegiventothehumanfactors,design,usability,and maintenanceofadevice.Testingmayinitiallytakeplaceinalaboratorysettingtoensurethe accommodationofinherentvariabilityindriveranthropometry,cognitiveandphysicalcapabilities, aswellasproperoperationwithinvariousdrivingenvironments.Furtherdesignandusability evaluationintheeldandlongitudinallywillaidinreningthefunctionalityofsuchtechnologies andassessingdriverdeviceusepatternsovertime. Additionally,givenvariousdevicestates,itiscriticaltodeterminethedegreetowhichdrivers understandthecapabilitiesandlimitationsofasystem,itsoperationalparameters,andwhatdriver actionsareexpectedinassortedsituations.Thedegreetowhichdevicesaccommodateindividual driversbypromotingcorrectinterpretationoftheiroutput,assumingindividualdifferencesinperception,informationprocessing,physical,andcognitiveskills,mustalsobeevaluated.Moreover, itisnecessarytoassessthedemandsofattendingtotheoutputofanin-vehicledevice,asasafety technologyshouldnevercontributetodriverstressorworkload.Asdevicefeedbacktypically takestheformofanauditorywarningoralert,itisvitalthatvariousoutputsareeasilycomprehended,discriminated,anddonotconictwiththoseprovidedbyothersafetytechnologies.Finally,inordertofacilitatetrustinthesafetybenetofsuchdevices,itiscrucialthatfalseand/or nuisancealarmsareminimized,inadditiontomaximizinghitsi.e.,correctdetectionofobjects intherearofthevehicle. 98

PAGE 115

Easeoflearningasapartofevaluatinguseracceptanceseeksinformationregardinghowwell adeviceisutilizedinitsintendedmanner,aswellaswhatisdonewithsuchacquiredknowledgeovertime.Basictestingofsuchparametersmaybeconductedinasimulatedsettingfor theevaluationofshort-termoutcomes.However,onlyalongitudinalstudyineldconditions allowsfortheassessmentoflearningovertime.Userunderstandingoftheapplicationsofdevice feedback,bothreactivelyandproactively,iscriticalforthesuccessofsuchtechnologies.The perceivedvalueelementofuseracceptanceassessesthedegreetowhichdriversperceiveasafer and/ormorealertdrivingenvironmentasafunctionofdeviceuse.Ideally,thedriverisableto utilizesuchsafety-enhancingtechnologiestofacilitatealertvehicleoperation,inconjunctionwith successfullyintegratingdevicefeedback.Anadditionalaspectofperceivedvalueisthedegree towhichdriversreportthattheseinnovationsenhancedrivingperformanceandsafetyintheir everydayoperations.Whenassessingthesecriteria,itisimportanttoalsoconsidertheundesirable outcomeofdrivers'inadvertentorpurposefulover-relianceonsuchtechnologiestodetectobjects inblindareas.Perceivedvaluemayalsobeimpactedasaresultofthedegreetowhichdriversunderstandandareinformedaboutdevicefunctioningandwhataspectsofdriverbehaviorthedevice helps.Forexample,ifrealorperceivedhealthrisksareassociatedwiththetechnology,driverswill weighsuchcostsagainstotherperceivedbenets.Additionally,usersmaybeconcernedabout datacondentialitytotheextentthatdevicesareusedtomonitor,store,andpossiblytransmit informationregardingtheirdrivingbehavior. Advocacyismeasuredintermsoftheextenttowhichdriversconsiderendorsingorpurchasinga safetydevice,anditisacriticalcomponentofuseracceptance.Ultimately,regardlessofapotentialsafetybenet,andeveninspiteofperceivedbenetsonthepartofthedriver,ifatechnology isnotattainablebytheintendedusers,itwillnotsucceedinthemarketplace.Therefore,thewillingnessofdriversbothtopurchaseasafetydevicewhetheronanindividualorcommercialbasis andtoendorseittoothersisavitalaspectofsuccessfuldeployment. 99

PAGE 116

7.3DeviceUsagePatterns Asshownsubsection6.3.1,adevicethatdemandsthevisualattentionofthedriverhastobetested thoroughlytoensurethattheattentionofthedriverfromthemaintaskofdrivingorkeepingthe vehicleundercontrolisnothinderedordecreaseddramaticallybecauseofthedevice.Alterations indriverbehaviormayoccurasafunctionofdeviceusageovertime.Ideally,thesechangesare intended,positive,andhaveapermanentimpactonsafevehicleoperationanddriverlifestyle. Evaluatingadriver'sallocationofcognitiveandtemporalresourcestomaintainsafedrivingserves toensurethatdrivingbehaviorisnotnegativelyaffectedbydevicesrequiringexcessivetimeand cognitiveresourcestomonitorandreactto.Ofadditionalimportanceisassessingthedegreeto whichdriverawarenessofandexposuretodevicefeedbackovertimeyieldsbehavioraladaptation.Examplesincludetheextenttowhichdeviceoutputisintegratedintodrivingbehaviorand thepotentialbenetsand/orrisksofusingatechnologyinanunintendedmanner.Further,user acceptanceshouldfocusonalterationsindrivingstylei.e.,habits,patternsthatarebroughtabout bymodicationstoglancepatternsasaresultofrespondingtodeviceoutput.Morebroadly,it isimportanttoassesswhetherextendedexposuretosuchsafetydevicesleadstooveralllifestyle changes. 100

PAGE 117

Chapter8:ConclusionsandFutureResearch 8.1Summary Camera/video-basedsystemshavebeenusedforseveralyearsassurrogatesformirrorsorasenhancementsinapplicationsincludingcommerciallargetrucks.Particularlysuccessfulhasbeenthe useofthesesystemswherealargeblindareaorNozoneexistsaroundthevehicle.Theseblind areasarelargerandmoreimportantinlargevehiclessuchastractor-trailertrucks.Specically theuseofsuchsystemsasrearviewvideosystemshasbeenincreasinglyseeninallvehicletypes inrecentyears.Thisdissertationexaminedhowsuchadevicecanhelpthedriverinavoiding potentialhazardsbehindthevehiclethatcouldleadtobackingcrashes.Themaindifferencewith previousresearcheffortswastocapturethedrivers'behaviorastheyusethesystem,andgainan understandingontheimportantfactorsinuencingitsuseandmaximumbenetgain.Inthecase oflargevehicles,arearviewofthedirectareabehindthevehicleisnotpossiblesincetheyhaveno rearwindow,rearviewmirrorortheirviewisblockedbythetrailer.Providingthedriverwithan additionalrearviewofthevehicleprovestobebenecialespeciallyincertainmaneuvers. TheFloridaCARsystemwasusedtoanalyzedbackingcrashdatafor2003-2006,togaininsight intothecharacteristicsofbackingcrashesinvolvinglargetrucks.Theproblemisnotassevereas othertypesofcrashesbutforlargetrucksthereisaconsistenttrendofabouttenpercentoftruck crashesbeingbackingcrashes.Themotivebehindthisresearchistoeliminateagreatpercentage ofthesecrashessinceitistheoreticallypossible.Thecrashanalysiswasalsodonefortwomain reasons.First,tounderstandthelocationsandconditionsunderwhichthecrasheshappen,and secondlytoestimatethepotentialcrasheswhichthedeviceisapplicable.Itwasfoundthatthe crashescodedasbackingcrashesinthedatabasemightnotnecessarilybebackingcrashes, 101

PAGE 118

ornotallbackingcrashesarecodedasbackingrelatedcrashes.Thiscreatedtheneedtoreview theactualpolicereportslongformtounderstandbetterwhathappenedandobtaininformation missingfromthedatabase.Surprisingly,backingcrashesoccurinlocationswherebackingisnot allowed.About20percentofbackingcrashesinvolvingtrucksoccurredatintersectionsornear intersections,another7percentoccurredatdrivewaysandabout26percentoccurredonaroad section.Onlyabout40percentoccurredatparkinglotsandprivatepropertywhereonewould expectthemtohappen.Atotalof1,549policereportswerereviewedtondthat0.17percent werefatal,7.85percentinvolvedinjuriesincapacitating,non-incapacitatingandpossibleinjury and89.2percentwerepropertydamageonly.Forthemajorityofthesebackingcrashes,more than70percent,theat-faultvehiclewasatruckandthesecondvehiclewasanautomobile.The differenceinvehiclesizeiswhatmakescrashesmoreseveresinceevenatlowspeeds,themassof thetruckcaninictgreatdamageandinjuries. Furthermore,theanalysiswasperformedtoestimatethenumberofbackingcrashesthattheRVS isapplicableto,thuscanbeclaimedtobepotentiallyavoidedoreliminated.Twoissueswere investigated.Thepointofimpactoftheat-faultvehiclesincetheRVScannothelpwithanyother locationonthevehicle,andvehiclespeeds.Itwasidentiedbypreviousresearchthatforthe systemtowork,bothvehiclesneedtobetravelingatlowspeedsbelow10mphandbeunder thecontrolofthedriver.ThiswaywhenthedriversperceivethedangerusingtheRVS,theyhave enoughtimetoreactandstopthevehiclebeforestrikingtheobjectorperson.Basedontheanalysiswithpointofimpactandspeedofvehiclesatthetimeofcrash,itseemsthat84percentoflow speedbothvehiclesbeloworat10mphbackingcrashesamong83percentofallbackingcrashes withrearsidepointofimpactcanbepotentiallyeliminatedoravoidedbyadoptingtheRVSasa countermeasure. Afterthecrashanalysis,thedriverbehaviorwhenusingtheRVSwasanalyzedthroughdatacollectedatacontrolleddriverexperiment.Theexperimentwasdesignedtotesttheeffectivenessof theRVSinhelpingthedriversdetectandavoidpotentialhazardsbehindthevehicleduringbackingmaneuvers.Aseriesofbackingmaneuverswereusedtoobtainsimulatedconictandcrash data.Atotalof45localtruckdriversparticipatedintheexperiment.Thedriversworkedforafood 102

PAGE 119

deliverycompany.Aday-cabtractoranda35ftboxtypetrailerwereusedfortheexperiment. ThreebackingmaneuversfoundintheCDLskilldrivingtestwereutilizedfortheexperiment.The driversperformedallmaneuverswiththeRVSandwithouttheRVSinarandomorder.Aremote controlledpedestriandummyandtrafcconewereusedassurrogatestocreatetheconicts.All thedriversweremaleswithaverageagerange31-40yearsold.Thedrivershadanaverage10.91 yearsofdrivingexperience,and15hadsomepreviousexperiencewiththeRVS.Also15drivers hadatleastonebackingcrashintheirhistory.Theexperimentyieldedatotalof71crashesout of270maneuvers.3%.Whenanalyzed,thethreemaneuversyieldeddifferentprobabilities ofhavingabackingcrashwithandwithouttheRVS.FortheStraightLineBackingManeuver therewasadecreaseof46.67percentinprobabilityofhavingabackingcrash,fortheOffsetRight BackingManeuvertherewasa4.44percentdecreaseinprobability,andfortheAlleyDockBackingManeuvertherewasadecreaseof17.78percent.Thisshowsthatthesystemhelps,butin varyingamountfrommaneuvertomaneuver.Alsotherstmaneuverhadadifferentintroduction ofobjectthantheothertwo. Driverbehaviorwasobservedduringtheexperimentandmeasuredforsignicantdifferences.The driverbehaviorwasseparatedintotwoaspects:rstwasthevisualattentionandsecondthedriver acceptanceofthesystem.Thevisualattentionmetricsincludedmeasuringwhereandhowlong thedriverwasglancingatdifferentlocations.Theselocationswereidentiedasdriversidemirror, passengersidemirror,RVSmonitorlocationandeverywhereelse.Thedriversneededonaverage 6.47secondsmoretimeforthemaneuverswiththeRVSinuse.Theyspentlesstimelookingat otherlocationsanddiditlessfrequentlyinordertoaccommodatetheadditionalglancelocation presentedtothem.Overalltheyseemedtobeabletomanagetheirtimewithsomeexceptions. Theseexceptionsincludedsomedriversthatdidnotusethemonitorasmuchoratallwhenitwas availabletothemandfocusedontheirusualroutine. Thedriveracceptanceofthedevicewasmeasuredwithasurveygiventothemaftertheycompletedthetest.Overallinallmeasuresthemajorityofdriversagreedthatthesystemhelpsin reducingtherearblindspotandthusitisahelpfuldeviceinreducingbackingcrashessinceitwill helpthemavoidpotentialhazardswhilebacking.Themajorityalsostatedthattheywouldliketo 103

PAGE 120

havethedeviceintheirtruckforeverydayoperations.Theseresultsshowanacceptanceofthe deviceandthereforethemaximizationofthedevice'suseandpotentialbenets. Themethodologyusedforthisstudycanbeimplementedinamoregeneralapproachforallinvehicledeviceevaluations.AlthoughprocedureshavebeenusedfromNHTSAfortheevaluation ofthesedevices,additionofthedriverusageaspectseemstobeimportantinordertoencompassa morerealisticpictureinimplementationofsuchdevices. 8.2ConclusionsandRecommendations Obviouslywiththedriverstillthemostimportantfactorindriving,theRVScanonlyprovide informationandhelpthedriveravoidhazards.Thisdeviceispassiveandprovidesnowarning tothedriver,ratheritpresentsarearviewmuchlikearearviewmirrorinapassengercarandthe driversneedstoobserve,detectandreactusingtheinformation.Fromthisanalysis,theRVSis successfulinreducingtherearblindspotandprovidesthedriverswithanadditionalrearview theydidnothavebefore.DuringtheexperimenttheRVSincreasedthestoppingrateofthedrivers, andreducedtheprobabilityofhavingabackingcrashundertheconditionsoftheexperiment.The participatingdriversalthoughskepticalinthebeginning,didshowsomechangeintheirperception ofthesystemaftertheyexperiencedthepotentialbenets. Thismethodofevaluatingasafetydeviceprovidesgreatbenetsforresearchers,especiallyinthe humanfactorsarea,wheretheexactmovement,usageandergonomicsofadeviceareanalyzedto obtainoptimumbenets.Usingcamerasandlivevideofeedwhiledrivingisacomplicatedtask andthedriversneedtogetusedtoittoobtainbetterresults.Somedriversinthisexperimentdid nothaveanypriorexperiencedwiththesystem,whereassomedidusethesystembeforealthough notforlongperiodsoftime.Asseenindriverbehaviorstudies,driverscanadapttheirdriving habitsandpatternstoencompassnewtechnologiesavailableinvehicleseveryyear.Severalissues havetobeaddressedbeforeadriverorcompanycanpurchaseandusesuchdevicesontheireet. Withtheincreaseinaftermarketsystems,customizationismoreexibleandcanproduceeven betterresultsthanOEMdeviceinterfaces.Theonlyissueisregulationofsuchsystemsbecause 104

PAGE 121

withouttestingandinspection,faulty,inappropriateandunsafesystemsmightemergeandwith timebecomepartoftheautomobileindustrysafetystandards. 8.3FutureResearch Futureresearchonthissubjectincludestestingwithadditionalcamerasinotherareasofthevehiclesinordertoidentifysolutionsforothertypesofcrashes.Alsotestingofdifferentlocationof thecameramightyielddifferentresults.Asanexampletheexperimentcouldbeperformedwith thecameralocatedatthetopofthecargobox.Sincetheviewisdifferent,itmightyielddifferent resultsthantheonespresentedhere.Alsoadifferenttypeoftruckmightbeusedtoevaluatethe deviceunderadifferentscenario. Inaddition,thetestingperformedforthisstudywaspossibleinashortperiodoftimeandwith limitedresources.Theevaluationofsuchdevicesinanaturalisticdrivingstudysuchastheone conductedin[25]canprovidebetterandmorerobustresultsindeterminingthetruebenetsofusingthesedevicesontheroad.Sinceeverydriver'shabitsaredifferent,observinginrealtimeover alongerperiodoftimecouldpresentwithadditionalissuesthatneedtobeaddressedtoensurea safeoperation. Also,theperception-reactiontimeofthedriverswiththeRVSandwithout,couldbeanalyzedto identifypotentialdifferences. 105

PAGE 122

ListofReferences [1]ORAE. EstimationofBackoverFatalities .ReportNHTSA-2006-25579-0001-2.NHTSA, 2006. [2]NHTSA. ReporttoCongress:VehicleBackoverAvoidanceTechnologyStudy .NHTSA,2006. [3]Wang,J.S.,andR.R.Knipling. BackingCrashes:ProblemSizeAssessmentandStatistical Description .TechnicalReportDOTHS808074.NHTSA,1994. [4]NHTSA. Non-TrafcDeathandInjuryDataCollectionStudy .TechnicalReportDOTHS 809-668.NHTSA,2004. [5]BureauofTransportationStatistics.FreightDataandStatisticsReports. http://www.bts.gov/programs/freight_transportation.Accessed:05/30/2009. [6]TRIP.TheImpactofIncreasedTruckingonEconomicDevelopment,CongestionandTrafc Safety.2004. [7]FDHSMV. TrafcCrashStatisticsReport .FloridaDepartmentofHighwaySafetyandMotor Vehicles,2005. [8] MichiganTrafcCrashFacts .MichiganOfceofHighwaySafetyPlanning,2005. [9] MinnesotaMotorVehicleCrashFacts .OfceofTrafcSafety,MinnesotaDepartmentof PublicSafety,2005. [10]WDOT. WisconsinTrafcCrashFacts .WisconsinDepartmentofTransportation,2004. [11]NCDOT.NorthCarolinaDepartmentofTransportationCrashDatabase,2008. 106

PAGE 123

[12] AnnualReportofFatalandInjuryMotorVehicleTrafcCollisions .StatewideIntegrated TrafcRecordsSystem,2005. [13]Tijerina,L.,D.Hendricks,J.Pierowicz,J.Everson,andS.Kiger. ExaminationofBacking CrashesandPotentialIVHSCountermeasures .FinalReportDOTHS808016.Battele, 1993. [14]Dene,L.J.,R.Grace,A.M.Guzman,D.A.Purta,andJ.G.Tabacchi. VisionEnhancement ResearchandTechnologyGuide .CarnegieMellonResearchInstitute,1994. [15]ITSAMERICA. CollisionAvoidanceSystems:IssuesandOpportunities .Virginia,1994. [16]Najm,W.G.Comparisonofalternativecrash-avoidancesensortechnologies. Intelligent VehicleHighwaySystems ,Vol.2344,1994,pp.62. [17]NHTSA. AStudyofCommercialMotorVehicleElectronics-BasedRearandSideObject DetectionSystems .FinalReportDOTHS808080,1994. [18]Garrott,W.R.,M.A.Flick,andE.N.Mazzae.HardwareEvaluationofHeavyTruckSideand RearObjectDetectionSystems. SAETechnicalPaperSeries ,1995. [19]Ruff,T.M. RecommendationsforTestingRadar-BasedCollisionWarningSystemsonHeavy Equipment .ReportofInvestigations9657,2002. [20]Ruff,T.M. EvaluationofSystemstoMonitorBlindAreasBehindTrucksUsedinRoad ConstructionandMaintenance .NIOSH,2003. [21]Mazzae,E.N.,andR.W.Garrott. ExperimentalEvaluationofthePerformanceofAvailable BackoverPreventionTechnologies .FinalReportDOTHS810634.NHTSA,2006. [22]Mazzae,E.N.,andR.W.Garrott. ExperimentalEvaluationofthePerformanceofAvailable BackoverPreventionTechnologiesforMediumStraightTrucks .FinalReportDOTHS810 865.NHTSA,2007. [23]Mazzae,E.N.,andR.W.Garrott.EvaluationofthePerformanceofAvailableBackover PreventionTechnologiesforLightVehicles.2007. 107

PAGE 124

[24]Lee,C.,andA.Kourtellis.CommercialTruckDriverSurveyforRearViewSystems.CUTR, 2008. [25]Mazzae,E.N.,F.Barickman,G.S.Baldwin,andT.Ranney. On-RoadStudyofDrivers'Use ofRearviewVideoSystemsORSDURVS .NHTSA,2008. [26]Weirwille,W.W.,W.A.Schaudt,S.K.Gupta,J.M.Spaulding,andR.J.Hanowski. DevelopmentofaPerformanceSpecicationforCamera/VideoImagingSystemsonHeavyVehiclesSpecications .VirginiaTechTransportationInstitute,2008. [27]Jenness,J.,R.Llaneras,andR.Huey. StaticTestingMethodforAssessingQualityofIndirect VisibilityonHeavyTrucks .Westat,Inc.,2008. [28]Weirwille,W.W.,W.A.Schaudt,S.K.Gupta,J.M.Spaulding,andR.J.Hanowski. DevelopmentofaPerformanceSpecicationforCameraVideoImagingSystemsonHeavyVehiclesSupportingResearch .VirginiaTechTransportationInstitute,2008. [29]NHTSA.FederalMotorVehicleSafetyStandards;RearviewMirrors;AdvanceNoticeof ProposedRulemaking,2009. [30]Johnson,D.E. AppliedMultivariateMethodsforDataAnalysts .DuxburyPress,1998. [31]Barr,L.,H.Howarth,S.Popkin,andR.J.Carroll. AReviewandEvaluationofemerging DriverFatigueDetectionMeasuresandTechnologies .NHTSA,2009. [32]Ball,D.,S.Versluis,R.Hendrickson,J.Pittenger,B.Frank,A.Stewart,andD.Murray. FactorsinDecisionstoMake,Purchase,andUseOn-boardSafetyTechnologies .Final ReportFMCSA-MCRT-06-003.Battelle,2005. [33]NHTSA. ReporttoCongressontheNationalHighwayTrafcSafetyAdministrationITS Program,ProgramProgressDuring1992-1996andStrategicPlanfor1997-2002 .U.S. DOT,1997. 108

PAGE 125

Appendices 109

PAGE 126

AppendixA:RearviewSystemComponentTechnicalSpecications TableA.1: 1 3 inchCCDrearviewcameraspecications. FeatureSpecications TVSystemEIA EffectivePixels512x492pixels SensingArea4.9mmx3.7mm ScanningSystem2:1Interlace Sync.SystemInternal Resolution420TVlines MinimumIllumination0Lux HorizontalSync.Resolution15,734KHz VerticalSync.Resolution60Hz VideoOutput1.0vp-p,75Ohm MicrophoneYes GammaConsumption0.45 AGCAuto S/NRatioBetterthan48dB WhiteBalanceAuto ElectronicShutter1 = 60 1 = 10 ; 000 BLCAuto CurrentConsumptionMax.300mA PowerSupply12VDC OperatingTemperature )]TJ/F19 10.9091 Tf 8.477 0 Td [(20 C 70 C ,RH95%MAX StorageTemperature )]TJ/F19 10.9091 Tf 8.477 0 Td [(40 C 85 C ,RH95%MAX LensAperturef=2.8 110

PAGE 127

AppendixA:continued TableA.2:LCD5inchmonitorspecications. FeatureSpecications DisplayDeviceColorTFT-LCD ScreenSize5inches AudioOutput200mW LoudspeakerOne4.0cmroundloudspeaker ConnectingTerminalEarphonejack,audio/videoAVinputjack,external powersupplyinputjack,rearviewconnector ApplicationPowerSupply12VDC PowerConsumption 8 W OuterDimensions163mmW,125mmH,30.5mmT Weight430g Resolution960H,234V ContrastRatio150:1 Brightness300 cd = m 2 111

PAGE 128

AppendixB:InformedConsentForm TitleofProject:YardTestsforRearviewVideoSystem Experimenters:Dr.ChanyoungLee,AchilleasKourtellis ThePurposeofthisResearch Fatalities,injuriesandpropertydamagecausedbybackingcrashhavereceivedmoreandmore attentionfromthepublic,truckingindustryandmanagementauthorities.Backingcrashesare usuallycausedbythe"No-Zone"wherethedriverhasvirtuallynovisibility.The"No-Zone" isespeciallysevereforheavytrucks.Currently,theCenterforUrbanTransportationResearch CUTRattheUniversityofSouthFloridaUSFisconductingresearchwiththeFloridaDepartmentofTransportationDistrict7toevaluatetheeffectivenessofRearview/Backupcamera systemsinreducingpotentialbackingcrashesamongcommercialvehicles. Asapartofthestudy,youarebeingaskedtoserveasaparticipant.Ifyouagreetoparticipate, youwilldriveaheavyvehicletractortrailerinthecompany'syard.Wewillgiveyoudetailed instructionsonwhattodolater,butbasicallyyouwillperformseveralstandardbackingmaneuvers.Youwillparticipatebyperformingthemaneuversinbaseline,thatis,withthevideoturned off,andalsowiththevideoturnedonsoyoucanuseit.Theorderinwhichthesewillbepresented isdifferentfordifferentparticipants.Youwillparticipateintheevaluationofacamerarearview systemandcorrespondingbaselineruns.Yourparticipationisexpectedtotakenomorethanone hour,butmaybeabitlongerorshorter. Procedures Hereinthebuildingyouwillrstdecideifyouwanttoparticipate.Ifso,youwillsignyourname attheendofthisform,soindicating.Youshouldonlysignafteryouhavereadandunderstoodthis formandhadyourquestionsanswered.Next,wewillgoovertheteststobeperformedandthe 112

PAGE 129

AppendixB:continued orderinwhichtheywillbepresentedtoyou.Foreachtypeofrun,youwillperformwhatwecall deliverydrivingtasks.Duringmostofthetime,youwillmaintainthenormalbackingspeed. Youwillbeaskedtoperformatotalofsixbackingmaneuver.Eachofthesewillbeexplainedin detailpriortohavingyouperformthemaneuvers.Ifyourvideoisturnedonduringthesetests, youshouldtrytouseitthemtoimproveyourperformance.Ofcourse,wedon'tknowhowwell theywillwork,soyourjobistojustdothebestyoucan.Wewilltakemeasurements,butthereis nograding,soyouwon'tpassorfail.Also,resultswillbekeptcondential,aswillbeexplained. RisksandDiscomforts Therisksyouwillfaceinthisexperimentareprobablyslightlylessthanyouwouldfaceindriving arigforyoureverydayjob.Speedsshouldbelow,andyoushoulddriveassafelyaspossibleoras youdriveeveryday.Consequently,webelievethisisaminimumriskexperiment.Wedon'tknow ofanydiscomfortsassociatedwiththeexperiment,exceptpossiblyyourworkingwithequipment youhaven'tusedbefore.Thismightcausealittlestress,butwethinkthestressshouldbemild. BenetsofthisProject Therearenodirectbenetstoyouforparticipatinginthisresearchotherthannormalparticipant payment.Nopromiseorguaranteeofbenetshasbeenmadetoencourageyoutoparticipate. Youmayndtheexperimentinteresting,andyourparticipationmayhelpintheevaluationofthis camerasystemonheavyvehicles. 113

PAGE 130

AppendixB:continued ExtentofAnonymityandCondentiality Theresultoftestwillremainanonymousandtherawdatawillnotbesharedbeyondresearchteam atCUTR.Also,whileyouaredriving,equipmentwillrecordvehiclepositionandsimilardata.In addition,wewillmakesomemeasurementsofvehiclenalpositionandsimilaraspectsforthe backingtasks.Inallcases,yournamewillbekeptseparatefromyourdata.Dataanalysiswillbe basedonthepooledresponsesofthosewhocompleteparticipation.Atthistime,itisanticipated thatatotalof60+driverswillparticipate.Itwillbeimpossibleinreportingtheresultsoftheexperimenttoidentifyanyparticularparticipant. Whileyoudriveinthisexperiment,yourglancepositionmayberecordedbyvideo.Thisisdone byaimingasmallvideo-cameraatyourface.Aftercompletionofyourparticipation,therecordingswillbeusedforresearchpurposesonlyandwillbeanalyzedtoextractyourglancepositions. Therecordingswillbekeptsecureuntiltheyarenolongerneeded.Theywillthenbeerased. Compensation Youwillreceivepaymentintheamountof$$perhourforyourtimeandparticipation.Thispaymentwillbemadethroughthecompanypleasenotethatexpectedprocessingtimeis8-12weeks. FreedomtoWithdraw Youshouldknowthatyouarefreetowithdrawfromtheexperimentatanytimeandforanyreason withoutpenalty.Noonewilltrytomakeyoucontinue.Ifyoudonotwanttocontinue,youwillbe paidfortheactualamountoftimeyouparticipated.Youarenotrequiredtoansweranyquestions ortorespondtoanyresearchsituations,andyouwillnotbepenalizedfornotresponding.The experimenteralsohastherighttoendtheexperiment,ifinhisopinionitisbesttodoso. 114

PAGE 131

AppendixB:continued Participant'sPermission IhavereadandunderstandtheInformedConsentandconditionsofthisproject.Ihavehadallmy questionsanswered.Iherebyacknowledgetheaboveandgivemyvoluntaryconsentforparticipationinthisproject.IfIparticipate,IunderstandthatImaywithdrawatanytimewithoutpenalty. NameDriverNumberforTestAssignedbyCUTR Participant'sSignatureDate 115

PAGE 132

AppendixC:AfterExperimentEvaluationForm DriverNumber: 1.DriverAge: AUnder25B-30C31to40D41to50E50+ 2.YearsofCDLdrivingexperience: yrs. 3.PreviousexperiencewithRearviewVideoSystem:AYESBNO IfYESabove:DurationofexperiencewithRVS: A1weekB1-2weeksC3-4weeksD1-2monthsE3-4monthsFNever 4.EverhadbackingcrashesduringCDLdriving?AYESBNO IfYESabove:Howmanybackingcrashes? 5.Basedonyourexperiencewithtruckingcompany,canyouguesswhatisthepercentageof backingcrashesrelativetoallcrashes? ALessthan10%B10%-20%C20%-30%D30%-40%E40%-50%FMorethan50% 6.WiththerearviewcamerasystemIfeelmorecomfortabletoperformthestraightbacking maneuver. 1StronglyAgree2SomewhatAgree3Agree4Disagree5SomewhatDisagree6Strongly Disagree 7.Therearviewcamerasystemhelpsmetoreducepotentialcrashesduringstraightbacking maneuver. 1StronglyAgree2SomewhatAgree3Agree4Disagree5SomewhatDisagree6Strongly Disagree 8.WiththerearviewcamerasystemIfeelmorecomfortabletoperformtheoffsetrightbackingmaneuver. 116

PAGE 133

AppendixC:continued 1StronglyAgree2SomewhatAgree3Agree4Disagree5SomewhatDisagree6Strongly Disagree 9.Therearviewcamerasystemhelpsmetoreducepotentialcrashesduringoffsetrightbackingmaneuver. 1StronglyAgree2SomewhatAgree3Agree4Disagree5SomewhatDisagree6Strongly Disagree 10.WiththerearviewcamerasystemIfeelmorecomfortabletoperformthealleydockbacking maneuver. 1StronglyAgree2SomewhatAgree3Agree4Disagree5SomewhatDisagree6Strongly Disagree 11.Therearviewcamerasystemhelpsmetoreducepotentialcrashesduringalleydockbacking maneuver. 1StronglyAgree2SomewhatAgree3Agree4Disagree5SomewhatDisagree6Strongly Disagree 12.Wouldyouliketohavethisrearviewcamerasysteminthetruckyoudriveeveryday? 1AbsolutelyNot2ProbablyNot3Not4Yes5ProbablyYes6AbsolutelyYes **IfyouansweredNOabovewhynot? 13.Ifthecompanyyouworkforisconsideringequippingtheeetwiththerearviewcamera systemtominimizebackingcrashesforthepriceof$300-$400persystem.Wouldyou supportthisdecision? 1StronglyAgree2SomewhatAgree3Agree4Disagree5SomewhatDisagree6Strongly Disagree 117

PAGE 134

AppendixC:continued 14.Doyoufeelthattherearviewcamerasystemhelpstoreducetherearblindspotofyour vehicle? 1StronglyAgree2SomewhatAgree3Agree4Disagree5SomewhatDisagree6Strongly Disagree 15.Arethereanyadditionalcommentsyouwouldliketomakeregardingtherearviewcamera system? 118

PAGE 135

AbouttheAuthor Mr.AchilleasKourtelliswasborninCyprus.HereceivedhisCivilEngineeringTechnicianDiploma attheHigherTechnicalInstituteinNicosia,Cyprusin2000.Histhesiswasonmixdesignmethodsofbituminousmixtures.Followingthatheservedamandatorytermof24monthsintheCyprus NationalGuardasasecondlieutenantinthedivisionofCivilCorpsofEngineers.In2002he transferredtotheCivilandEnvironmentalEngineeringDepartmentattheUniversityofSouth Floridawherein2004hegraduatedwithhisB.Sc.inCivilEngineeringwithconcentrationin Transportation.Hewasimmediatelyadmittedtothegraduateprogramandin2006hegraduated withaMastersinCivilEngineeringwithconcentrationinTransportation.Hethencontinuedtothe Ph.D.program.HeworkedattheCenterforUrbanTransportationResearchatUSFasagraduate researchassistantsince2004wherethersttwoyearshewasworkingintheMobilityPolicyProgram,andthelatter3yearsintheITS,TrafcOperationsandSafetyProgram.Heisco-authorof 3papersandrstauthorof3conferencepapersinthesubjectpresentedinthisdocument.Hehas alsoreviewedpapers,andpreparedseveraltechnicalreports,presentationsandotherdocuments.


xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam 2200421Ka 4500
controlfield tag 001 002069523
005 20101007115922.0
007 cr mnu|||uuuuu
008 100422s2009 flu s 000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0003252
035
(OCoLC)608554857
040
FHM
c FHM
049
FHMM
090
TA145 (Online)
1 100
Kourtellis, Achilleas.
0 245
Operational evaluation of advanced safety enhancement devices :
b rearview video system
h [electronic resource] /
by Achilleas Kourtellis.
260
[Tampa, Fla] :
University of South Florida,
2009.
500
Title from PDF of title page.
Document formatted into pages; contains 118 pages.
Includes vita.
502
Dissertation (Ph.D.)--University of South Florida, 2009.
504
Includes bibliographical references.
516
Text (Electronic dissertation) in PDF format.
520
ABSTRACT: Since the creation of the automobile, there has been an effort to create and implement mechanical and electronic devices that would improve vehicle safety. In recent years, electronic technologies have become more efficient and cost effective, therefore creating a great spike in widespread implementation. These safety related devices have to be tested for their reliability and amount of help they provide the driver with. The end user (the driver) has to be involved for a successful device. This research presents the methodology used to evaluate the effectiveness of the rearview video system (RVS) used in vehicles, especially in large commercial trucks and effectively the methodology for a more complete investigation of the problem of correctly implementing a safety device. The focus of this research is backing crashes that involve large trucks. The countermeasure tested was a rearview video system which provides a rear view to the driver in real time.A traditional crash data analysis is almost impossible since there is not enough data to perform it, and no data are available for the use of this system since it is fairly new to the market. A driver experiment under controlled conditions was used to create and collect the data necessary for the analysis. The experiment yielded a total of 71 crashes out of 270 maneuvers (26.3%). When analyzed, three backing neuvers yielded different probabilities of having a backing crash with and without the RVS. The increase in stop rate ranged from 46.67 percent to 4.44 percent. This is interpreted as crash reduction due to the device. Driver behavior was observed during the experiment and measured for significant differences. The drivers needed on average 6.47 seconds more time for the maneuvers with the RVS in use. They spent less time looking at mirrors and did it less frequently in order to accommodate the additional glance location presented to them.Overall they seemed to be able to manage their time with some exceptions. The driver acceptance of the device was also measured with a survey given to them after they completed the test. Overall in all measures the majority of drivers agreed that the system helps in reducing the rear blind spot and thus it is a helpful device in reducing backing crashes since it will help them avoid potential hazards while backing. The majority also stated that they would like to have the device in their truck for every day operations. These results show an acceptance of the device and therefore the maximization of the device's use and potential benefits. The RVS is therefore effective in reducing potential backing crashes. The results presented here are limited, and inferences are made with the experiment conditions in mind. General application of the results is possible, with certain assumptions and restrictions.
538
Mode of access: World Wide Web.
System requirements: World Wide Web browser and PDF reader.
590
Co-advisor: Jian (John) Lu, Ph.D.
Co-advisor: Pei-Sung Lin, Ph.D.
653
Camera
Methodology
Implementation
Experimental
In-vehicle
Driver behavior
690
Dissertations, Academic
z USF
x Civil Engineering
Doctoral.
773
t USF Electronic Theses and Dissertations.
4 856
u http://digital.lib.usf.edu/?e14.3252