自動(dòng)駕駛汽車整車在環(huán)仿真測(cè)試平臺(tái)-2023-03-自動(dòng)駕駛

第二屆車路協(xié)同自動(dòng)駕駛論壇(The2ndForumofCAV-HighwaySystems)弘毅明德篤學(xué)創(chuàng)新Prof.ZhigangXu/deputydirectorInternetofvehicleLabsponsoredbyMinistryofEducationandChinaMobileChang’anUniversity.長(zhǎng)安大學(xué)車聯(lián)網(wǎng)教育部-中國(guó)移動(dòng)車聯(lián)網(wǎng)實(shí)驗(yàn)室徐志剛副主任/教授njing弘毅明德篤學(xué)創(chuàng)新WhyweneedtotesttheautonomousvehicleWhatcanwelearnfromexistingexperiences現(xiàn)有測(cè)試方法研究3.TheproposedHILSimulationtestplatformandongoingwork智能汽車硬件在環(huán)測(cè)試平臺(tái)4.Opportunitiesandchallenges機(jī)遇與挑戰(zhàn)弘毅明德篤學(xué)創(chuàng)新1、Whyweneedtotesttheautonomousvehicle為什么要進(jìn)行智能汽車測(cè)試弘毅明德篤學(xué)創(chuàng)新Thedevelopmentofautonomousvehicles(AV)caneffectivelyreducetrafficaccidents&energyconsumption,improvetrafficefficiency,andfreethedriverfromboringdrivingtasks.弘毅明德篤學(xué)創(chuàng)新1.2Accidents&SafetyItisimminenttoestablishacompletetest&evaluationframeworkforicalscenariosforAVse弘毅明德篤學(xué)創(chuàng)新tsTest12.5yearsTestyears1.09deaths/100millionmilesaccidentrateaccidentrateaccuracy:95%performanceofAVexceeds20%ofHVThedrivingenvironmentishighlyuncertain,non-repeatable,unpredictableandinexhaustibleeportoftheRANDCorporationstatesTraditionalautomotivesafetytestingmethodshavebeenisnecessarytostudynewtestingtheoriesanddesignnewtestingmethods.HowtorealizetherapidtestofAVs,improvingtestefficiency&reducingtest弘毅明德篤學(xué)創(chuàng)新2、Whatcanwelearnfromexistingexperiences現(xiàn)有測(cè)試工具研究弘毅明德篤學(xué)創(chuàng)新ISO26262.Repeatable、Extreme&DangerousConditions、Accelerate弘毅明德篤學(xué)創(chuàng)新Consistsofsimulationscenarios,vehicledynamicsmodel,sensormodel,planningdecisionalgorithm,etc.Getridoftheneedforrealtestenvironmentandhardware.Generallyappliedintheearlystagesoffunctionaldevelopment.Highefficiency&Lowcost、risk&ShortperiodThetestresultsareheavilydependentonthethesensormodelandthevehiclemodel.Rapidrestorationofreal-worldscenariosenvironmentisdifficulttoachievecorrectnessofinsimulatedAirSimApollooo弘毅明德篤學(xué)創(chuàng)新FieldtestFromtheenvironmenttothevehiclesarephysical.Flexibleintegrationofscenariosandenvironments.Multiplescenariosandtestcases.LowefficiencyLongperiod 弘毅明德篤學(xué)創(chuàng)新Somepartsorsystemsarereal,andtheenvironmentisvirtual.Acceleratethetestingprocess.Improvethetestefficiency.Realistictestresults.SeveralHILtestingtechniquesarerelativelycomplexandcostly,suchasLiDAR,infraredcameras,etc.Thescopeofapplicationisstilllimited弘毅明德篤學(xué)創(chuàng)新ationtestplatformandongoingwork智能汽車硬件在環(huán)測(cè)試平臺(tái)弘毅明德篤學(xué)創(chuàng)新ChanganUniversitymainlyconductsfunctional&performancetesting,safety&reliabilitytesting,andachievescomprehensiverapidvalidationofsafety,sindoorsgTestSite弘毅明德篤學(xué)創(chuàng)新mVirtualSimulationTestScenarioe弘毅明德篤學(xué)創(chuàng)新mFrontAxleRotatableDrumTestBenchSimulaterealroadphysicalenvironmentindoors:CurvedsimulationwithmultidegreeoffreedomarchitectureturedynamicsimulationPavementadhesioncoefficientadjustment(snowyroad,slipperyroad)hanglerollangleandyawangleCollectdatasuchasspeed,wheelangleandhorizontal/longitudinalslope弘毅明德篤學(xué)創(chuàng)新mSensorintheLoopSimulationMillimeterWaveRadarintheLoopSimulation:ionofVirtualMMWRadarionRealMMWRadarionGeometricmodelestimationPhysicalmodelRadarmodelGeometricmodelestimationPhysicalmodelRadarmodelTargetmodelEnvironmentmodel弘毅明德篤學(xué)創(chuàng)新mSensorintheLoopSimulationMillimeterWaveRadarintheLoopSimulation:ting弘毅明德篤學(xué)創(chuàng)新mSensorintheLoopSimulationMillimeterWaveRadarintheLoopSimulation: CarCameraCarCarCameraCarCameraVirtualTrafficScenarioCarCamera弘毅明德篤學(xué)創(chuàng)新mSensorintheLoopSimulationCameraintheLoopSimulation:CameraofAVSchemeofCameraofAVDataDataProcessingUnitDataDataProcessingUnitVirtVirtualCameraDataDataProcessingUnit弘毅明德篤學(xué)創(chuàng)新mSensorintheLoopSimulationCameraintheLoopSimulation:CameraObscuraSchemeofCameraCameraObscuraVirtualVirtualTrafficScenarioVirtVirtualCameraCameraCameraofAVaalHDCamera弘毅明德篤學(xué)創(chuàng)新mSensorintheLoopSimulationCameraintheLoopSimulation:ramSensorintheLoopSimulationLiDARintheLoopSimulation:CollectedPointCloudinRealScenarioGenerated3DScenarioGeneratedPointCloudbySimulatedLiDARintCloudbySimulatedLiDARontestontest弘毅明德篤學(xué)創(chuàng)新mTestSituationonditionroad車硬件在環(huán)仿真測(cè)試平臺(tái)弘毅明德篤學(xué)創(chuàng)新ionoftestingenvironment令ReconstructionofLiDARdata令reconstructionofCameradata令ReconstructionofMMWRadardata令ConstructionofHigh-precisionmap令Synchronizationofdatafvehiclesenvironment令Roadmodeling令Vehicledynamicsandkinematics令Pedestrian,obstaclesimulation令Simulationofdangerousscenes令Pedestrian-Vehicle-Roadcoordination令Quantificationofintelligence令Descriptionofthesemanticnetwork令Man-machinegame令Learningabilitytest令Endurancetest車硬件在環(huán)仿真測(cè)試平臺(tái)弘毅明德篤學(xué)創(chuàng)新andchallenges總結(jié)與挑戰(zhàn)Conclusions1.AVILworkbenchintegratingvehicle,simulatedroadsystemsandsensorswasestablished,whichcanbeusedtoverifythefunctionsandperformanceofanautonomousvehicle.開(kāi)發(fā)了一種集成車輛、虛擬場(chǎng)景、模擬傳感器的室內(nèi)無(wú)人車快速測(cè)試臺(tái)，可以用于無(wú)人車功能驗(yàn)證和性能測(cè)試。2.Threetypesofsimulatedsensormodels(MMWRadar,RGBCameraandLiDAR)areconstructed,whichcanfeedtheperceptiondataofthesimulatedscenariostothedecisionalgorithmofthetestedautonomousvehicle.構(gòu)建了三種傳感器仿真模型(毫米波、RGB彩色攝像頭、激光雷達(dá))，可以為無(wú)人的決策算法測(cè)試提供多種類型的感知數(shù)據(jù)。3.SomeexperimentsonsimplescenariossuchasACC,AEBandLKPhavebesuccessfullyconductedontheproposedworkbench,whichisprovedthattheproposedwor