Perpustakaan Digital ITB

The yellowsignaldriverbehavior,reflectingthedilemmazonebehavior,isanalyzedusingnaturalistic data fromdigitalenforcementcameras.Thekeyvariableintheanalysisistheentrancetimeafterthe yellow onset,anditsdistribution.Thisdistributioncanassistindeterminingtwocriticaloutcomes:the safety outcomerelatedtored-light-runningangleaccidents,andtheefficiencyoutcome.Theconnection to otherapproachesforevaluatingtheyellowsignaldriverbehaviorisalsodiscussed. The datasetwasobtainedfrom37digitalenforcementcamerasatnon-urbansignalizedintersections in Israel,overaperiodofnearlytwoyears.Thedatacontainmorethan200millionvehicleentrances, of which2.3%(?5 millionvehicles)enteredtheintersectionduringtheyellowphase.Inallnon-urban signalized intersectionsinIsraelthegreenphaseendswith3sofflashinggreen,followedby3sofyellow. In mostnon-urbansignalizedroadsinIsraelthepostedspeedlimitis90km/h. Our analysisfocusesoncrossingsduringtheyellowphaseandthefirst1.5softheredphase.The analysis methodconsistsoftwostages.Inthefirststagewetestedwhetherthefrequencyofcrossingsis constant atthebeginningoftheyellowphase.Wefoundthatthepatternwasstable(i.e.,thefrequencies were constant)at18intersections,nearlystableat13intersectionsandunstableat6intersections.In addition tothe6intersectionswithunstablepatterns,twootheroutlyingintersectionswereexcluded from subsequentanalysis.Logisticregressionmodelswerefittedforeachoftheremaining29intersection. Weexaminedbothstandard(exponential)logisticregressionandfourparameterslogisticregression.The results showaclearadvantagefortheformer.Theestimatedparametersshowthatthetimewhenthe frequency ofcrossingreducestohalfrangesfrom1.7to2.3safteryellowonset.Thedurationofthe reduction oftherelativefrequencyfrom0.9to0.1rangedfrom1.9to2.9s.