以人为中心的汽车主动安全预警信息系统研究

重庆大学博士学位论文以人为中心的汽车主动安全预警信息系统研究姓名：廖传锦申请学位级别：博士专业：控制理论与控制工程指导教师：黄席樾20050501I————————————IIIIIABSTRACTTheresearchingontechnologyofvehicleactivesafetyisintheascendantnow.TheR&Doftechnologyofvehicleactivesafety,suchascollisionavoidance,toreducethedriver’sburdenormisjudgmentisimportanttoimprovethetrafficsafety.Indespiteofthelargedevotionofenergyonautomaticdrivingandtheresultoftheresearching,thereisnooneaccuratemodeltodescribethedrivingprocessveritablyandcompletely,becausethedrivingprocessisahighintelligentizedprocess.So,it’sclearthatforewarningistheeffectivetechnicalmeanstoimprovethedrivingsafety.Thegenesisanddevelopmentofvehicleactivesafetytechnology,theideaofhuman-centeredvehicleactivesafetyandthestatesofartsofvehicleactivesafetytechnologyareintroducedbrieflyinthisdissertation.Inallusiontotheresearchingstatus,oneideaispresentedthatallthevehicleactivesafetytechnologyshouldbehuman-centered,i.e.driver’ssafety-centered,driver’sperceiveidentity-centered,driver’soperationidentity-centered.Thecharacteristicandflowofinformationandthethreephasesofdriving,i.e.perception,decision-making,operation,areanalyzedinthisdissertation.Asimplemodelofvehicle-driver-environmentispresentedbasedonanalyzingthereciprocityofvehicle,driverandenvironmentintheclosedloop.Thefunctionalstructureofhuman-centeredvehicleactivesafetyforewarninginformationsystemisformedandthemodelofforewarningorientedhuman-centeredvehicle-driver-environmentispresentedtomeettheneedofsafetyforewarning.Afteranalyzingthetraditionalmodelofhumaninformationperceptionandprocessingbasedonattentionsingleresourcetheory,oneconceiveofdynamicresourceandstaticresourceofattentionisformedbasedontheattentionmulti-resourcetheory.Thedistributingmodelsofdynamicresourceandstaticresourceofattentionarepresented,andbasedonthesemodels,anewmodelofhumanordriverinformationperceptionandprocessingispresented.Inallusiontothenewmodel’sfunctionalphases,somemodelsareformed,i.e.,themodelofmatchingdegreeofcharacterindexesbasedonfuzzymeasurablefunction,thepriorityweightdistributingmodelofinformationcharacterindexesbasedonHAP,theinformationsortclassifyingmodelbasedonsubjectiondegree,theinformationimportanceexponentbasedoninformationsortsubjectiondegree,sortpriorityweightandhumanorientationcoefficient.ThemodelofIVinformationselectionbasedontheimportanceexponentispresented.Basedonanalyzingthecharacteristicsandshortagesoftraditionaldecision-makingmodels,anewdecision-makingmodelwithself-learningmechanismispresentedaimingatthebasiccharacteristicsofdrivingtask.Somemodelsorschemearepresentedforthisnewmodelofdecision-making,i.e.thefeaturematchingmodelbasedonevidencefusion,correspondenceanalysismodelbasedonweightedEuclideandistance,drivingstatesafetyevaluationmodelbasedonfuzzycenterofgravity,refusalcoefficienttorealizetheselectionofdrivingoperationscheme,theself-learningmodelbasedonmappingtransposition.Driver’ssafetyconsciousnessmeasureisintroducedtothisdissertationaccordingtotheroleofsafetyconsciousnessinthedrivingtask.Astatisticmodelofsafetyconsciousnessmeasureispresented.Driver’ssafetyconsciousnesswillbeweightedtoauditthedrivinghabitsreal-timebythemodel.Amodeloffatigueaccumulationbasedonfatigueelementispresentedandthefeaturefatigueinthephasesofperception,decision-makingandoperationisanalyzedwiththeresponsecurveofunitrampfunction.Analgorithmoffatigueanalyzingbasedonsafetyconsciousnessmeasureispresented.Thehuman-centeredforewarninginformationsystemforvehicleactivesafetyisintroducedfromthreeaspects,i.e.informationcapture,riskevaluation,structureofsystemhardware.Informationcaptureisintroducedfromroadinformation,hostvehicle,obstacleinformation.Animprovedoptimalthresholdalgorithmforimagesequencesispresentedtorealizethelanesegment.Analgorithmofimagespeedsensorbasedonlanestateperiodicispresentedtomeasurethespeedofhostvehiclereal-timeandexactly.Aconceptofgraylevelcountsofunitroadsurfacepixelisinductedandbasedonit,analgorithmofobstaclelocationisdesigned.Thealgorithmofmonocularmeasurementisadoptedinthisdissertationandthefilterofdistancedataisrealizedwiththetheoryofentropy.Basedoninformationcaptureandanalyzingaclassicdrivingstate,analgorithmofriskevaluationordecision-makingbasedonminimalsafetydistanceispresented.Keywords:Vehicle,ActiveSafety,Forewarning,Human-centered1111.120211.11986200401020304050607080901986198719881989199019911992199319941995199619971998199920002001200220032004Year1.119862004Fig.1.1somestat.dataofroadtrafficaccidentfrom1986to2004inChina280%65%11.2[1]2080ABSBASEBSASR13ITS19861470120Prometheus[2]901990ITSMobility200021IVSH301997[3]2080ABSBASEBSASR1.31.293%[4]2080HowardRosenbrok[5]42090ASVActiveSafetyVehicle[6],1.3457227276311.2%Fig.1.2theproportionofinducementsoftrafficaccidents(%)151.4[7]1.4.14[8]1.11.14Table1.1thecontrastof4sensorsinvehiclenavigationtechnique1.3Fig.1.3thefunctionstructureofhuman-centeredactivesafetysystem6Navlab198919912070CMU2080[919]CMUPALPHRapidlyAdaptingLateralPositionHandlerALVINNAutonomousLandVehicleinaNeuralNetSCARFSupervisedClassificationAppliedtoRoadFollowingUNSCARFUNSupervis