交通管理系統(tǒng)中的安全性：一項全面調(diào)查 Safety in Traffic Management Systems A Comprehensive Survey

/journal/designs

Designs2023,7,100.

/10.3390/designs7040100

[eess.SY]11Aug2023

Review

SafetyinTrafficManagementSystems:AComprehensiveSurveyWenluDu,AnkanDash,JingLi,HuaWeiandGuilingWang*

Citation:Du,W.;Dash,A.;Li,J.;Wei,H.;Wang,G.SafetyinTraffic

ManagementSystems:AComprehensiveSurvey.Designs2023,7,100.

/10.3390/

designs7040100

arXiv:2308.06204v1

AcademicEditor:RobertoGabbrielli

Received:4June2023

Revised:1July2023

Accepted:13July2023

Published:10August2023

Copyright:?2023bytheauthors.LicenseeMDPI,Basel,Switzerland.Thisarticleisanopenaccessarticledistributedunderthetermsand

conditionsoftheCreativeCommons

Attribution(CCBY)license

(https://

/licenses/by/

4.0/).

YingWuCollegeofComputing,NewJerseyInstituteofTechnology,Newark,NJ07102,USA;wd48@(W.D.);ad892@(A.D.);jingli@(J.L.);hua.wei@(H.W.)

*Correspondence:gwang@

Abstract:Trafficmanagementsystemsplayavitalroleinensuringsafeandefficienttransportationonroads.However,theuseofadvancedtechnologiesintrafficmanagementsystemshasintroducednewsafetychallenges.Therefore,itisimportanttoensurethesafetyofthesesystemstopreventaccidentsandminimizetheirimpactonroadusers.Inthissurvey,weprovideacomprehensivereviewoftheliteratureonsafetyintrafficmanagementsystems.Specifically,wediscussthedifferentsafetyissuesthatariseintrafficmanagementsystems,thecurrentstateofresearchonsafetyinthesesystems,andthetechniquesandmethodsproposedtoensurethesafetyofthesesystems.Wealsoidentifythelimitationsoftheexistingresearchandsuggestfutureresearchdirections.

Keywords:survey;trafficsafety;proactivesafetymethods;safetyanalysis;crashprediction;crashriskassessment;deeplearning;machinelearning;statisticalanalysismethods

1.Introduction

AsaddressedbytheU.S.DepartmentofTransportationStrategicPlanFY2022–2026(

/dot-strategic-plan

)(accessedon14March2023),makingthetransportationsystemsaferforallpeopleisstillatopstrategicgoal.About95%oftransportationfatalitiesintheUSAoccuronthecountry’sstreets,roads,andhighways,andthenumberofdeathsisincreasing.Trafficsafetyisofparamountimportance,particularlyintheeraofemergingtechnologieslikeautomatedvehiclesandconnectedvehicles[

1

].Asthesetechnologiescontinuetoevolveandbecomemoreprevalentontheroads,thepotentialforsafertransportationincreasessignificantly.Automatedvehicleshavethepotentialtominimizehumanerror,whichisresponsibleforthemajorityoftrafficaccidents.Withtheiradvancedsensorsandalgorithms,theycandetectandrespondtopotentialhazardsmoreswiftlyandeffectivelythanhumandrivers.Similarly,connectedvehiclesenablereal-timecommunicationbetweenvehiclesandinfrastructure,allowingforenhancedawarenessandcoordinationontheroad.Thisconnectivityfacilitatestheexchangeofcriticalinformation,suchastrafficconditions,weatherupdates,androadhazards,therebyenablingdriverstomakeinformeddecisionsandavoidpotentialdangers.Byembracingandprioritizingtrafficsafetyinconjunctionwiththeseadvancedtechnologies,wecanstrivetowardsafuturewithreducedaccidents,injuries,andfatalitiesontheroadways,ultimatelycreatingasaferandmoreefficienttransportationsystemforall.

Inthepastfiveyears,researchershavemadesignificanteffortsinthefieldoftrafficsafety[

2

–

4

].Someresearchers,particularlythoseincivilengineering,havefocusedonstatisticalanalysistoidentifyandprioritizecountermeasures.Byanalyzinghistoricaldatasetsandrecords,theyaimtounderstandthecause-and-effectrelationshipsanddevelopeffectivestrategies.Forinstance,theyexaminecontributoryfactorssuchasadverseweatherconditionsthatincreasetheriskofaccidents[

5

–

7

].Thisanalysishelpsindevisingcontrolplans,includingdriverwarnings,toreducecrashratesduringextremeweatherevents.Ontheotherhand,interdisciplinaryresearchersaimtoprovideaccurateriskinformationbyutilizingmachinelearninganddeeplearningmodels[

4

].Theyworktowardsdevelopingreal-timecrashriskpredictionsystems.However,whenitcomestooperationalaspects,

2of30

Designs2023,7,100

lessattentionhasbeengiventoexplainingtheimpactofvariablesandmoreemphasishasbeenplacedonimprovingpredictionaccuracy.Techniqueslikedeepneuralnetworks,generativemodels,reinforcementlearning,ensemblemethodslikeXGBoost,andcomputervision-basedalgorithmshavegainedpopularityinthisregard.Thissurveyprovidesanoverviewoftheadvancementsinthesetwodirections,summarizingthestate-of-the-artresearchinthefield.

Furthermore,wehavecategorizedtherecentliteraturebasedonthespecificareasofanalysisorcontrol.Putsimply,someresearchersconcentrateonenhancingtheoverallsafetyofanentiretrafficnetworkoraspecificregion[

8

,

9

],suchasdowntownNewYorkCity.Othersaddresscrash-relatedissuesoccurringonhighwaysegments[

10

],onramp/offrampsections[

11

],weavingareas[

12

],andcurvedsegments[

13

].Additionally,effortshavebeenmadetoimprovesafetyatintersections[

14

].Asautomatedvehicles,connectedvehicles,andconnectedandautonomousvehicle(CAV)technologycontinuetoemerge,alongwithadvancedfeatureslikeautomaticemergencybraking(AEB)withpedestriandetection,adaptivecruisecontrol(ACC)systems,andadvanceddriverassistancesystems(ADAS),studieshavefocusedonthevehiclesideaswell[

3

].Forinstance,researchersevaluatethereal-timeriskofcollisionsinscenariosinvolvingcarfollowing[

15

]orplatooning[

16

].Inthissurvey,wealsoprovideanoverviewofexistingresearchintheseaspects.

Finally,weconcludebyaddressingthepresentchallengesandlimitations,aimingtoprovideaclearunderstandingofareasthatcanbeimprovedinthefuture.Throughourcomprehensiveliteraturereview,weobservedthatcertainlimitationsareprevalentandremainunresolvedtothisday.Onesuchchallengeistheimbalanceddataproblem[

17

],whichsignificantlycomplicatespredictivetasksduetothelimitedrepresentationofcrashdatawithinthedataset.Manyresearchershighlightthedifficultyincollectinglabeledacci-dentdatainreal-worldscenarios[

10

].Anothercommonissueisthelackofgeneralizabilitytoreal-worldconditions[

18

],assomeproposedmodelsdemonstratesatisfactoryperfor-manceonlyinsimulatedenvironments,withlimitedevidenceofsuccessfuldeploymentinreal-worldsettings.Itisessentialtorecognizeandaddressthesechallengesinordertoadvancethefieldoftrafficsafetyandimprovetheapplicabilityoftheresearchfindingsinpracticalcontexts.

Itisimportanttonotethedivergencebetweentheresearchconductedinthefieldofcivilengineeringandinterdisciplinaryresearch,particularlyincomputerscience.Civilen-gineeringresearchersoftenemploystatisticalanalysisandsensitivityanalysistoexplorethecorrelationbetweenvariablesandtheirimpactonsafety.Theyfrequentlyutilizereal-worlddatasetsandconductfieldteststoobtainempiricalevidence.Conversely,interdisciplinaryresearcherstendtoprioritizethedesignofmodelsusingsimulatedenvironments,whichmayormaynottranslateeffectivelyintopracticalapplications.However,thereisagrowingtrendtowardsintegratingdomain-specificruleswithpopularneuralnetworkmodelstoleveragethestrengthsofbothapproaches[

19

].Thiscollaborativeapproachaimstobridgethegapandcapitalizeonthebenefitsofferedbycombiningdomainknowledgewiththecapabilitiesofneuralnetworks.

Thesurveypapermakesthefollowingcontributions:

?Athoroughexaminationoftheliteraturepublishedwithinthelastfiveyearsiscon-ducted,allowingforanaccuratedepictionoftheprevailingresearchtrendsduringthisperiod.

?Theliteraturecollectionexclusivelyfocusesontop-tiervenues,ensuringthattheselectedworksarehighlyrepresentativeofboththedomainfieldandcomputersciencefield.Thisprovidesvaluableinsightsforresearchersinterestedintrafficsafetyapplications.

?Wecategorizetheworksintotwodistinctcategoriesofanalysisandcontrolandprovidecorrespondingsummariesthatoutlinetheresearchobjectivesandlimitations.Thiscategorizationoffersinspirationandguidanceforfutureresearchersinthefield.

3of30

Designs2023,7,100

2.ReviewMethod

Inthissection,weoutlinetheprocessofcollectingourreviewpapers.Weprovidedetailsregardingthenumberofpapersreviewed,theirrespectivesources,andtherelevantstatistics.Forinstance,wehighlightthedistributionofpapersbetweenthecomputerscienceandtransportationfields,sheddinglightontherepresentationfromeachdiscipline. Toinitiateourpapercollection,weconductedkeywordsearchesfortermssuchasroadsafety,accidentprevention,accidentavoidance,crashrisk,andtrafficaccidentacrosstop-tiervenuesinboththetraditionaltransportationfieldandcomputerscienceandrelateddisciplines.Thesurveycoversaspanoffiveyears,specificallyfrom1January2019to

1June2023.Figure

1

presentsagraphicalrepresentationofthepublicationdistributionduringthisfive-yearperiod,distinguishingbetweenpublicationsinthetransportationfieldandthoseinthecomputersciencefield.Whilecomputerscienceencompassesdiversere-searchareas,wediscoveredseveralrenownedvenueswherecomputerscientistscontributetheirwork,applyingproposedmodelstothetransportationdomainanddemonstratingtheirpracticality.Althoughthenumberofpublicationsincomputerscienceisrelativelysmallcomparedtothatinthetransportationfield,thereisanevidentupwardtrendinpublicationsovertheyears,indicatingthepressingneedtoenhancesafetymeasures.Thisupwardtrajectorysuggeststhattheriseinpublicationswilllikelycontinueinthefuture.

Figure1.Collectedpaperpublicationsindifferentfieldswithinthepastfiveyears.

Figure

2

displaysacomprehensivelistofthetop-tiervenuesutilizedinthissurvey,alongwiththecorrespondingpublicationcountsforeachvenue.Inthetransportationfield,weobservedthatIEEETransactionsonIntelligentTransportationSystemsaccumulatedthehighestnumberofpublications,suggestingitsprominenceamongresearchersfordisseminatingtheirwork.Additionally,conferencesinroboticengineeringalsocontributedseveralpublications,withaprimaryfocusonautonomousdrivingandrelatedtechniques.

Figure2.Collectedpaperpublicationsbypublicationvenueswithinthepastfiveyears.

4of30

Designs2023,7,100

3.SurveyingtheLiterature:AnIn-DepthExploration

Basedontheanalysisofthesurveyedpapers,wecategorizedthemintofourdistinctsectionsasshowninTables

1

and

2

.Firstly,wehighlighttherecurringproblemsortopicsthatfrequentlyappearedinthepublications.Secondly,weprovideasummaryoftheareaswheresafetyimprovementsareemphasized,suchasintersectionsorfreeways.Thirdly,weexaminethespecifictargetsthatresearchersfocusedonintheireffortstoenhancesafety,suchasconnectedvehicles,andwecompileacomprehensivelistofthetrendingtechniques

discussedinthesurveyedpapers.

Table1.Keywordsummary.

Topics

?

?

?

?

?

?

?

?

IdentificationofDangerousVehiclesAccidentForecasting

IdentificationofCrashRisk

CrashRiskAssessment

Real-timeProactiveRoadSafety

ForwardCollisionAvoidance

Rear-endCollisionAvoidanceSecondaryCrashLikelihoodPrediction

?

?

?

?

?

?

?

?

CyclistCrashRatesAssessment

RareEventModeling

Inter-vehicleCrashRiskAnalysisIdentificationofHigh-riskLocationsTrajectoryPredictions

TrajectoryCollisionAvoidancePredictivePlatoonControlPedestrianOccupancyForecasting

?

Spatial–temporalCorrelations

?

Signal-vehicleCoupledControl

?

Minute-Level

?

CarFollowing

?

DriverBrakingBehavior

?

Take-overPerformance

?

Driver’sEvasiveBehavior

?

Left-turnatSignalizedIntersections

?

Heavy-truckRisk

?

Safety-awareAdaptiveCruiseControl

?

MovingVehicleGroups

?

AdaptiveMergingControl

?

School-agedChildren

?

LaneKeepingSystem

?

Evacuation

?

In-vehicleWarning

?

OldDrivers

?

Context-aware

?

SocialVulnerability

?

Multitask

?

DrivingImpairmentsandDistractions

?

HumanDriverImitation

?

PedestrianCrashRiskAnalysis

?

DashcamVideos

?

AutomaticEmergencyBrakingSystems

?

DriverDrowsinessMonitoring

?

Precipitation

?

HazyWeatherConditions

?

SurrogateSafetyMetrics

?

On-rampMergingControl

?

AdaptiveTrafficSignalControl

?

PreferencesofAggressiveness

Uponcarefulobservation,weidentifiedcrashriskpredictionasthemostextensivelyaddressedtopicamongthesurveyedpapers.Itoccupiedasignificantportionofthelitera-turereviewed.Furthermore,wenoticedagrowingtrendoffocusingonspecificconditionsorscenarios,suchasheavy-truckrisk,school-agedchildren,evacuation,extremeweather,andmore.Thesepapersaimedtoaddresssafetyissueswithinthesespecificsituationsandproposemeasuresforimprovement.Withtheadventofadvancedtechnologies,safetyconcernsrequirereassessmentandreevaluation.Someworksdelvedintothesafetyimpli-cationsofemergingtechnologies,suchasanalyzingtake-overperformanceorexaminingtheimpactofautomaticemergencybrakingsystemsonoverallsafety.Additionally,cer-tainhigh-riskareasthatfrequentlyexperienceaccidentshavegarneredattentionfromresearchers,leadingtofocusedinvestigationsontopicslikeon-rampmergingcontrol.Moreover,newsurrogatesafetymetricshaveemergedashighly-discussedsubjectswithintheliterature,furtherreflectingtheshiftinglandscapeoftrafficsafetyresearch.

Itisimportanttoacknowledgethedisparitybetweentheresearchconductedinthetransportationdomainandtheresearchpursuedbycomputerscientists.Thetraditionaldomainapproachesprimarilyfocusedonanalyzingthecontributoryfactorsleadingtocrashes,whereascomputerscienceresearcherswereinclinedtowardsdesigningmoreeffectivemodelsforriskpredictionandsafetyplanning.Inthesubsequentsections,we

5of30

Designs2023,7,100

adheretothislogicaldistinctionbyfirstdelvingintotheanalysisofthecontributoryfactorsandsubsequentlyintroducingvariouscontrolmethods.

Table2.Keywordsummarybydifferentperspectives.

InvestigatedLocations

?FreewaySegments

?HorizontalCurvature

?Expressways

?Intersection

?RoadwaySegments

?UrbanArterials

?TypeAWeavingSegments

?RingRoads

ConsideredEntities

?ConnectedVehicles

?AutonomousVehicles

?Cycling

?Motorists

?Pedestrians

Techniques

?BayesianNetwork

?DeepReinforcementLearning

?ReinforcementLearningTree

?InverseReinforcementLearning

?ComputerVision

?MatchedCaseControl

?PropensityScore

?SHapleyAdditiveExPlanation

?GradientBoosting

?LSTM-CNN

?TransferLearning

?AttentionNetwork

?SupportVectorMachines

?StackedAutoencoder

?GatedRecurrentUnit

?MonteCarloTreeSearch

?ImitationLearning

Data

?

?

?

NaturalisticDrivingData

SimulatedData

DrivingSimulatorPlatform

??

SHRP2NDSData

Event-basedData

3.1.OngoingFundedResearchProjects

Inadditiontothemajorscientificconferencesandjounals,wealsoinvestigateactivefundedresearchprojectsonsafetyincludingNCHRP,FHWA,andNHTSA,aimingtosummarizethelatesttrendinpractice.

NCHRPRsearchProjects.AU.S.researchprogramaddressingtransportationchal-lenges,administeredbyTRBundertheNationalAcademies,NCHRPfundsprojectsonvarioustopics,includinghighwaysafety,involvingexpertsfromacademia,industry,andgovernmenttoenhancetransportationsafety.NCHRPprojectsaimtoenhancetrafficsafetyanddevelopstrategiesforpedestrians,bicyclists,androadinfrastructure.Theycoverareassuchastrafficsafetyculture,pedestriansafety,highway–railgradecrossings,ruralhighways,alternativeintersections,motoristbehavior,andleveragingAIandbigdata.Theresearchfocusesonimprovingsafety,reducingcrashes,andprovidingdecision-makingtoolsfortransportationdepartments.Specifically,NCHRP17-96aimstodevelopapriori-tizedresearchroadmapforTrafficSafetyCulture(TSC)toimprovetrafficsafetybychangingvaluesandattitudesandstrategicallyapplyingTSCstrategiesincollaborationwiththe4Es.NCHRP17-97investigatesthecausesofnighttimepedestriancrashes,evaluatestheexistingandemergingstrategiesforimprovingpedestriannighttimesafety,proposeseffec-tivemitigationstrategies,anddevelopsguidancefortheirimplementation.NCHRP17-99developsaframeworkandtoolsforassessingthesafetyeffectivenessoftreatmentsandtechnologiesathighway–railgradecrossings,aidingdecisionmakingtoreduceincidentsandimprovesafety.NCHRP17-92developsapredictivemethodologyforestimatingthecrashfrequencyandseverityonruraltwo-lanetwo-wayhighways,incorporatingspeedmeasures.NCHRP17-109developsCrashModificationFactors(CMFs)forAutomatedTrafficSignalPerformanceMeasures(ATSPM)signaltiming,quantifyingsafetybenefitsandcrashreductionsforallmodesandconflicttypes.NCHRP17-108developsquantitativecrashpredictionmethodologies,includingSafetyPerformanceFunctions(SPFs)andCrash

6of30

Designs2023,7,100

ModificationFactors(CMFs),foralternativeintersectiondesigns(DLT,MUT,andRCUT)toquantifytheirsafetybenefits.NCHRP17-106quantifiestheeffectsofcenterlineandshoulderrumblestripsonbicyclists’safetyandworkstounderstandmotorists’behavior,informingdesignpoliciesanddevelopingaguideforrumblestripapplications.NCHRP17-100leveragesAI,machinelearning,andbigdatatoprovidedata-drivenanalysistoolsandprioritizeinvestmentsforsaferroads,focusingonpedestrians,cyclists,andnew-mobilityusers.

FHWAResearchProjects.TheFHWAisaU.S.governmentagencythatmanagesandimprovesthecountry’shighwaystomakesuretheyaresafe,efficient,andaccessible.Theyworkonprojectsrelatedtoroadinfrastructure,trafficmanagement,andtransporta-tionplanning,playingacrucialroleinmaintainingandenhancingthetransportationnetworkforpeopleandgoods.FHWA-PROJ-19-0014aimstodevelopanArtificialRealisticData(ARD)generatorforevaluatingsafetyanalysismethods.FHWA-PROJ-20-0030linksdatabasestodevelopspeed-relatedCrashModificationFactors(CMFs)forsafetyanalysis.FHWA-PROJ-21-0069usesAImodelstopredicttrafficconditionsandmanagehighwaysproactively.FHWA-PROJ-20-0054createsasafetyassessmenttoolforinterchangedesigns.FHWA-PROJ-19-0089focusesonhumanfactorsinautomatedvehicles.FHWA-PROJ-19-0085evaluatesintersectiondesignsforpedestrianandbicyclistsafety.FHWA-PROJ-19-0026collectsdataandevaluatessafetyimprovementsformini-roundabouts,wrong-waydriving,andbicycleintersections.FHWA-PROJ-20-0002studiesthesafetyofpedestriancrossingsignswithLEDs.

NHTSAResearchProjects.TheNHTSA,aU.S.federalagencyundertheDepartmentofTransportation,activelypromoteshighwaysafety,setsvehiclestandards,andreducestrafficinjuries.ItfacilitatestheESVconference,aplatformforsharingresearchandinitiativesonvehiclesafety,withpaperspublishedintheTrafficInjuryPreventionJournal.Afterreviewingtherecentpublications,wesummarizedthefollowingstudies:

Astudy[

20

]investigatedtheimpactofsexonfatalityratesincarcrashes,findingthatnewervehiclesandadvancedsafetyfeatureshavereducedfatalityrisksforfemaleoccupantscomparedtomales.Anotherstudy[

21

]evaluatedoccupantmodelswithactivemusclesandshowedtheirabilitytoaccuratelypredictoccupantresponsesincrashsimu-lations.Aninvestigation[

22

]focusedonelderlyindividualsinnear-sideimpactcrashesrevealedtheneedforfurtheranalysisinestablishinginjurythresholds.Astudy[

23

]ondrowsy-drivingdetectionmodelsincorporatedmultipledatasourcesandachievedgoodaccuracyinpredictingdrowsiness.Astudy[

24

]evaluatedthecrashreductionsachievedincarsequippedwithautomaticemergencybraking(AEB)systemswithpedestrianandbicyclistdetection.Theanalysisshowedanoverallreductioninthecrashrisk,withAEBsystemsreducingthepedestriancrashriskby18%andthebicyclistcrashriskby23%duringdaylightandtwilightconditions.However,nosignificantreductionswereobservedindarkness.Anothermethod[

25

]wasdevelopedtoaccuratelyandefficientlysimulatevehiclecollisions,providingcollisionseverityparametersforinjurymitigationassessment.Regulationsarebeingdevelopedforthesafeintroductionofautomateddrivingsystems,andadata-drivenscenario-basedassessmentmethodwasproposed[

26

]toestimatetheirsafetyrisk.

Throughourinvestigation,weobservedatrendtowardsutilizingadvancedtechnolo-gies,suchasactivemuscles,AEBsystems,anddata-drivenmodels,toenhancesafetyinvariousaspectsofcarcrashes.Sex-specificanalysisandunderstandingtheimpactofsexonfatalityrateshavegainedattention.Accurateprediction,detection,andassessmentofrisksarecrucialforenhancingsafetymeasures.Ongoingeffortsfocusondevelopingtechnologiesandmethodsforsimulatingandassessingcollisionseverity,aimingtoenhanceinjurymitigationcapabilities.

7of30

Designs2023,7,100

3.2.GeographicalDistributionoftheStudyArea

Weexaminedtheresearchlocationshighlightedintherecentliterature,specificallythesiteswheretheirexperimentswereconducted,asdepictedinFigure

3

.Ouranalysisre-vealedthatFlorida,USA,andShanghai,China,emergedastwocommonlychosenlocations.

US

China

Canada

Korea Japan Indian Europe Netherland Brazil

Figure3.Spatialdistributionofstudyregion:Variedcolorsdepictdiversecountries,andgreatercirclesizesignifiesmoreextensiveresearchinthatlocation.

4.Analysis

Theanalysisofsafetyintrafficmanagementsystemsinvolvesevaluatingandassessingthesafetyaspectsofvariouscomponentsandprocesseswithinatransportationsystem.Itaimstoidentifythepotentialhazards,assesstherisks,andimplementmeasurestomitigatethoserisks,ultimatelyensuringthesafetyofroadusersandminimizingtheoccurrenceofaccidents.Inadditiontoriskanalysis,researchersalsostrivetoanalyzeinjurieswiththegoalofminimizingtheiroccurrenceandseveritytothelowestpossiblelevel.Theanalysisofsafetyintrafficmanagementsystemsisamultidisciplinaryfieldthatcombinesexpertisefromtransportationengineering,dataanalysis,humanfactors,andpolicymakingtoensuresaferroadenvironmentsandreducethelikelihoodandseverityofaccidentsandinjuries.

4.1.Method

Wesummarizethemethodsusedfortheanalysisoftrafficsafety.

Matched-pairAnalysis.Matched-pairanalysis,alsoknownaspairedanalysisorpairedcomparison,isastatisticalmethodusedtocomparetworelatedsetsofdataorobservations.Itisparticularlyusefulwhenstudyingsituationswhereitisdifficulttoestablishadirectcause-and-effectrelationshipbetweenvariablesorwhendealingwithdatathatexhibitahighdegreeofvariability.Inmatched-pairanalysis,eachobservationinonegrouporconditionispairedormatchedwithacorrespondingobservationintheothergrouporcondition.Thepairingisconductedbasedonsimilaritiesorrelevantcharacteristicsbetweentheobservations,suchasage,sex,orsomeotherrelevantfactor.Thepairingensuresthateachpairofobservationsisassimilaraspossible,exceptforthevariablebeinginvestigated.Bypairingobservations,ithelpstocontrolforindividualdifferencesorconfoundingvariablesthatcouldaffecttheoutcomebeingmeasured.Thisanalysismethodincreasestheprecisionandreducesthepotentialbiasesassociatedwithunpairedcomparisons.Matched-pairanalysiswasappliedin[

6

]toanalyzetherelativecrashriskduringvarioustypesofprecipitation(rain,snow,sleet,andfreezingrain).

MutualInformationTheory.Mutualinformationtheoryisaconceptininformationtheorythatmeasurestheamountofinformationthatissharedortransmittedbetweentworandomvariables.Itquantifiesthedegreeofdependenceorassociationbetweenthevariablesandprovidesameasureofthereductioninuncertaintyaboutonevariablegivenknowledgeoftheothervariable.Entropyisafundamentalconceptininformationtheorythatcharacterizestheuncertaintyorrandomnessofarandomvariable.Itmeasurestheaverageamountofinformationneededtospecifytheoutcomeofarandomvariable.Higher

8of30

Designs2023,7,100

entropyindicateshigheruncertainty.Inaddition,mutualinformationmeasurestheamountofinformationthattworandomvariablesshare.Itquantifiesthereductioninuncertaintyaboutonevariablebyknowingthevalueoftheothervariable.Mathematically,itisthedifferencebetweentheentropyoftheindividualvariablesandthejointentropyofthetwovariables.Ifthemutualinformationishigh,itindicatesastrongrelationshipbetweenthevariables,suggestingthatknowledgeofonevariableprovidessubstantialinformationabouttheothervariable.Overall,mutualinformationtheoryhasproventobeavaluabletoolinvariousdisciplinesthatdealwithdataanalysisandinformationprocessing.Usingmutualinformationtheory,onestudy[

27

]quantifiedtheinteractionsbetweenvariousriskfactors,consideringmultifactorscenarios.

MatchedCaseControl.Thematchedcase-controlapproachcanbeappliedtoanalyzecrashoccurrencesduringspecialscenariossuchasevacuations[

7

,

28

].Thisapproachallowsforathoroughinvestigationofthepotentialriskfactorsorexposuresthatcontributetocrashesinaspecialscenariowhilecontrollingfortheconfoundingvariables.Forexample,theauthorsin[

7

],discussedastudyfocusedonunderstandingthefacto