自動駕駛電動助力轉(zhuǎn)向自動化外文文獻翻譯、中英文翻譯、外文翻譯

英文文獻原文ElectricPowerSteeringAutomationforAutonomousDrivingJ.E.Naranjo,C.González,R.García,T.dePedroInstitutodeAutomáticaIndustrial(CSIC)Ctra.CampoRealKm.0,200,LaPoveda,ArgandadelRey,Madrid28500,Spain{jnaranjo,gonzalez,ricardo,tere}@iai.csic.esAbstract.Theautomaticcontrolofavehicle’ssteeringwheelisnowoneofthemostimportantchallengesintheIntelligentTransportationSystemsfield.Inthispaper,wepresentafuzzylogic-basedautomaticsteeringcontrolsystemformass-producedelectricpowersteering(EPS)wheel-equippedvehiclesthatassureshuman-likebehavior.Intheliterature,wefindalotoftheoreticalproposalsandsomesimulations,butonlyafewworkteamsofferrealsolutionsforthistask.OnesuchsolutionistheworkdevelopedbytheAutopiaPrograminwhichsomevehicleshavebeenautomatedandcanperformsomemaneuversmimickinghumanreactions.Inthispaper,weusetheEPSofaCitro?nC3Plurieltoeffectthecar’sbehavior.Theactuatoriscontrolledfromanonboardcomputerhousingafuzzylogic-basedautonomoussteeringsystem.Thevehicle’sinternalcomputersgeneratetheinputinformation,whichisreadbyaCANbusandahighprecisionGPS.Someexperimentsusingthisequipmentonaprivatetestcircuitarepresented,obtaininganhuman-likebehaviorinallthemaneuvers.1IntroductionThedevelopmentofIntelligentTransportationSystems(ITS)providesanopportunitytoapplyadvancedtechnologytosystemsandmethodsoftransportforefficient,comfortableandsafermeansoftransport.Ourworkfocusesontheareaofroadtransport,andmorespecificallyonthefieldofintelligentvehicles,whichincludesthetopicofautonomousvehicles.Thistopicreferstovehiclesthatareequippedwiththeinstrumentationandintelligenceneededtoprovidetheactualvehiclewiththerequiredservice,thatis,anautonomouscarmustcontrolsomeorallofitsfunctionswithoutexternalintervention.TheAutopiaProgramisworkingonthisfield,focusingmainlyonautonomousdrivingusingfuzzylogiccontrollers.Thesteeringwheel[1],throttle[2]andbrakepedal[3]havebeenautomated,workinginCitro?nBerlingovansandexperimentswererunonprivatecircuits.Therearesomeexamplesofautomaticsteeringwheelcontrol,asasteptowardsachievingautomaticdriving.Inthe“MillemigliainAutomaticaTour”[4],acarwasequippedwithaDCmotorattachedtothesteeringwheelthroughapulley,whichcouldbemoveddependingonthecommandsofanonboardcomputerhousingananalyticalcontrolsystemthatreceivedthesensorinputthroughartificialvision.IntheAutopiaProgram[5],twoCitro?nBerlingovanshavebeenequippedforautomaticdriving.Here,thesteeringwheelhasalsobeenmodifiedtobemovedbyaDCmotorusinggearsattachedtothesteeringbarbecausetheassistmechanismofconventionalcarsispoweredbyhydraulicsystems.Inthiscase,thesensorinputisprovidedbyaGlobalNavigationSatelliteSystem.Newervehiclesareequippedwithelectricpowersteering(EPS),wheresteeringisassistedbyanelectricmotorthatactsdirectlyontherackbarthroughapinion.Themotortorquedependsontheeffortrequiredbythedriver.Theadvantageofthiskindofpowersteeringforourpurposesisthatnoexternalactuatorhastobeadded,andwecaneasilymanagetheassistmotorfromouronboardcomputer.Somedevelopments,forexample,Toyota’sautomaticparkingsystemhavetakenthisapproach[6].Theaimofthispaperistopresentthefuzzylogic-basedEPScontrolsystemdevelopedforautomaticdrivingthathasbeeninstalledinaCitro?nC3Plurielvehicle(Fig.1).Thisvehiclehasbeentestedonaprivatecircuit,andhuman-likebehaviorshavebeenachieved.Fig.1.Citro?nC3Plurieltestbedvehicle.2OnboardEquipmentElectricpowersteeringbasicallyconsistsofatorquesensorandmotoractuatorcouple.Thesensorisattachedtothesteeringcolumnandmeasuresthetorqueappliedbythedriverwhenhemovesthesteeringwheel.Thistorquesignalistransmittedtoacontrol/powercardthatsendsanamplifiedproportionalpowersignaltotheDCmotor,whichisengagedtothesteeringrackbar.Thefirststepforachievingautomaticsteeringcontrolistomanagethewheelsfromacomputerthatwehaveinstalledinthecar.Themethodforthisautomationistobypassthesensorandcontrol/powercardequipmentandsendapowersignaldirectlytothemotor.Ouronboardcomputerrunsafuzzylogic-basedcontrolsystemthatgeneratesacontrolanalogsignal.AnexternalpowerdrivehasbeenaddedforsupplantingtheoriginalC3’spowercard.Itusesasinputtheanalogsignalproducedbythecomputerandtheoutputisapowersignalthatsuppliestheassistmotor.Thevehicleinstrumentationiscompletedwiththesensorequipment:acarrierphasedifferentialGPSreceiver,whichgeneratesto-the-centimeteraccuratepositioning,andaCANbusinterface.NowwecanmoveontodescribethecontrolsystemdevelopedtomanagetheassistDCmotorand,consequently,thesteeringofthevehicle.3SteeringControlSystemAtwo-layerfuzzycontrollerhasbeendefinedforsteeringcontrol.Thehigh-levellayercalculatesthetargetpositionofthesteeringwheeltofitthevehicletothedesiredroute.ThelowlevellayergeneratestheoptimumtorquethatmustbeexertedbytheEPSassistmotortomovethesteeringwheelinahuman-likeway.ThereisalsoacomputationalreferencetrajectoryrepresentationrepresentedbythesetofthemostrepresentativeGPSwaypointsoftheroutetobetracked[1].3.1SteeringPositioncontrollerTwovariablesareusedasinputforthefuzzysteeringpositioncontrolsystem,namelylateralandangularerrors.Wedefinelateralerrorasthedistancebetweenthefrontofthevehicleandthereferencetrajectorysegment,measuredalongalineperpendiculartothatreferencesegment.Similarly,angularerrorisdescribedastheangleformedbythereferencesegmentofthetrajectoryandthecar’sdirectorvector.Wedefinetwofuzzyvariables,alsonamedangularerrorandlateralerror,eachofwhichhastwolinguisticlabels,leftandright,thatindicatewherethevehicleislocatedwithrespecttothereferencesegment.Bothvariableshaveoneassociatedmembershipfunctionforeachlabel,definedbytheirvertex,asshowninFig.2aandFig.2b.Fig.2.InputVariablesMembershipFunctions:a)andb)steeringpositioncontroller;c),d)ande)steeringtorquecontroller.Theoutputofthesystemisthetargetturninganglethatthesteeringwheelmustbemovedtocorrectthetrajectorydeviationindicatedbytheinputvariables.Thereisonlyonefuzzyoutputvariable,namedSteering,withtwolinguisticlabels,calledleftandright,whosemembershipfunctionsaredefinedbysingletonsTherulesetforgeneratingthesteeringturninganglefromtheinputdatabelowisthesameforbothbendandstraight-roaddriving,aswellasforfuzzycontrolandhumandriving.Thequalitativeactionsforthehumandriver(rules)arethesameinbothcases,andonlythequantitativepartvaries,whichisdefinedinthefuzzycontrolbythefuzzificationofthevariables:R1.1:IFLateral_ErrorLeftTHENSteeringRightR1.2:IFLateral_ErrorRightTHENSteeringLeftR1.3:IFAngular_ErrorLeftTHENSteeringRightR1.4:IFAngular_ErrorRightTHENSteeringLeftWherethewordsinitalicsarethefuzzyvariables,theonestotheleftofthetermTHENbeinginputvariablesandthevariablestotherightbeingoutputvariables.Thewordsinnormaltypearethelinguisticlabelsassociatedwitheachoneofthefuzzyvariables.3.2SteeringTorquecontrollerInthiscase,threeinputvariablesareneededtocontrolthetorqueappliedtothesteeringwheel.Thefirstistheangularpositionerrorofthesteeringwheel,thatis,thedifferencebetweenthetargetpositiongeneratedbythehighlevelfuzzycontrollerandtherealposition.Thesecondinputvariableistherealpositionofthesteeringwheel,andthelastoneistheangularspeedatwhichthesteeringwheelisturning.Whenthesevariablesarefuzzifiedforuseinthefuzzycontroller,theyaretransformedintofuzzyvariablescalledAng_Speedfortheangularspeed,Pos_Errorfortheangularpositionerror,Pos_Absfortherealsteeringwheelpositionand,respectively,associatedwiththemembershipfunctionsshowninFig.2c,d,ande.Theoutputofthefuzzycontrollerindicatesthevoltagethatmustbesenttothemotorpowercardthatappliesaproportionalamperagetothemotortomovethesteeringwheelwiththeoptimumtorquetocorrectlyachieveitstargetposition.Twolinguisticlabelshavebeendefined,Positive(right)andNegative(left),whosemembershipfunctionshavebeendefinedassingletons.Thedefinitionoftherulesaccountsfortheinteractionbetweentheinputandoutputvariablesthatwillgeneratetheoptimumcontrollerbehavior.Inthiscase,wehavedefinedsixrulesforcontrollingtheappliedtorque.R2.1:IFPos_ErrorPos_LargeTHENTorquePositiveR2.2:IFPos_VolNeg_LargeTHENTorqueNegativeR2.3:IFPos_AbsNegativeANDPos_ErrorNeg_SmallTHENTorqueNegativeR2.4:IFPos_AbsPositiveANDPos_ErrorPos_SmallTHENTorquePositiveR2.5:IFAng_SpeedMORETHANNullTHENTorquePositiveR2.6:IFAng_SpeedLESSTHANNullTHENTorqueNegative4ExperimentsHavinginstalledthedescribedcontrollerintheinstrumentedtestbedcar,weransomeautomaticsteeringcontrolexperiments,oneofwhichisshowninFig.3.Inthisfigure,theblackdottedlinerepresentsthereferencetrajectoryandthegraylineistheautomaticvehicleroute.Thisstartsatthecoordinates459028.75mNorth4462552.09mEast,behindthestartingpointlabel,andiscomposedofeightturns,fourtotheleftandfourtotheright,separatedbystraightsegmentsFig.3.Automaticroutetrace.Thecontrollersusetheinputvehicletrajectoryvariablesand,fromthisinformation,calculatethenecessarytorquetomovethesteeringwheel.Fig.4includesatraceofcontrollerbehaviorwhiletakingthefirstbendtothelefttoshowhowthesystemworks.Thetopgraphshowstheinputvariablevaluesfortakingthebend.Thenextgraphplotstheoutputofthesteeringpositionfuzzycontroller.Thethirdgraphcontainsthevaluesoftheinputvariablesofthetorquefuzzycontroller.Finally,thebottomgraphshowstheoutputtorque,normalizedfrom-1to1,tobeappliedtotheEPSmotorcontrollingthesteeringwheel.Atthebeginning,thecarisdrivingcenteredalongthefirstreferencesegmentoftheroute.Then,aseriesofnewpointsareloaded,andthelateralerrorincreasestotheleftandtheangularerroraugmentstotheright.Thisisnormalbehaviorsincethenewreferencesegmentstendtobeperpendiculartothefirstsegment,becausetheanglebetweentheconsecutivestreetsisabout90o.Theseinputvaluesarefuzzified,andtheruleinferenceofthehigh-levelcontrollerisexecuted,generatingaleftturningcommandthatisillustratedbythesteeringoutputvariable.Theoutputofthelow-levelcontroller(torque)showsthatthemaximumeffortisappliedatthebeginningoftheturning,whenapeakisneededtoinitiatethesteeringmovement(1-2sec).Oncethemovementisunderway,thetorquedecreasesrapidly.Finally,thecontrollermaintainsthesteeringpositionandmovesthesteeringwheelbacktothecenterwhentheturninghasfinished.5ConclusionsInthispaper,wehavepresentedatwo-layerfuzzycontrollerforautomaticelectricpowersteeringcontrol,whichwehaveusedtorunanumberofautomaticdrivingexperimentsdiscussedinthelastsection.Theseresultsshowedthatelectromechanicalsystems,likeanEPS,canbemanagedinahuman-likewayusingartificialintelligencetechniques,inthiscase,fuzzylogic.Thismethodallowstheusertomimichumanbehaviorbyextractingknowledgefromexperts,inthiscase,drivers.Anadditionaladvantageoffuzzylogicisthatcomplexnonlinearvehiclemodelsdonotneedtobedeveloped.Fig.4.Detailofthecontrolinputandoutputvariablesforthefirstturningtotheleftoftheautomatictrackingexperiment.ReferencesR.Garcíaetal.,“FrontalandLateralControlforUnmannedVehiclesinUrbanTracks”,IEEEIntelligentVehiclesSymposium,Versailles,France,2002.JE.Naranjoet.al.,“AdaptiveFuzzyControlforInter-VehicleGapKeeping”,IEEETrans.ITS,SpecialIssueonACC,Volume4,No.3,September2003,pp.132-142.J.E.Naranjoetal.,“AThrottle&BrakeFuzzyController:TowardstheAutomaticCar”,LNCS2809,Springer-Verlag,July2003,pp291-301.A:Broggietal.,“TheARGOAutonomousVehiclesVisionandControlSystems”,InternationalJournalofInt.Cont.andSyst.,Vol.3,No.4,pp.409-441,1999.J.E.Naranjoetal.,“OvertakingManeuverExperimentswithAutonomousVehicles”,Proc.oftheICAR2003,Coimbra,Portugal,pp.1699-1703,June2003.T.Endoetal.,“Developmentofreverseparkingassistwithautomaticsteering”,ITS2003CongressProceedings,Madrid,November2003.英文文獻譯文自動駕駛電動助力轉(zhuǎn)向自動化摘要：車輛方向盤的自動控制現(xiàn)在是智能交通系統(tǒng)領(lǐng)域最重要的挑戰(zhàn)之一。在本文中，我們提出了一種基于模糊邏輯的自動轉(zhuǎn)向控制系統(tǒng)，用于批量生產(chǎn)的電動助力轉(zhuǎn)向（EPS）輪式裝載車輛，以確保類似人的行為。在文獻中，我們發(fā)現(xiàn)了很多理論建議和一些模擬，但只有少數(shù)工作組為此任務(wù)提供了真正的解決方案。一個這樣的解決方案是由Autopia計劃開發(fā)的工作，其中一些車輛已經(jīng)被自動化并且可以執(zhí)行模擬人類反應(yīng)的一些操縱。在本文中，我們使用Citro?nC3Pluriel的EPS來實現(xiàn)汽車的行為。執(zhí)行機構(gòu)由一個機載計算機控制，內(nèi)置一個基于模糊邏輯的自主轉(zhuǎn)向系統(tǒng)。車內(nèi)的計算機生成輸入信息，由CAN總線和高精度GPS讀取。提出了在私有測試電路上使用這種設(shè)備的一些實驗，在所有操作中獲得類似人的行為。1引言智能交通系統(tǒng)（ITS）的發(fā)展提供了一種機會，將先進技術(shù)應(yīng)用于系統(tǒng)和運輸方式，以實現(xiàn)高效，可靠和安全的運輸方式。我們的工作重點是公路運輸領(lǐng)域，特別是智能車輛領(lǐng)域，其中包括自主車輛的主題。本主題是指配備有為實際車輛提供所需服務(wù)所需的儀表和智能的車輛，即自主車必須控制其部分或全部功能而無需外部干預(yù)。Autopia計劃正在開展這一領(lǐng)域，主要側(cè)重于使用模糊邏輯控制器的自主駕駛。方向盤[1]，油門[2]和制動踏板[3]已經(jīng)自動化，在雪鐵龍Berlingo廂式車上工作，實驗運行在專用電路上。有一些自動方向盤控制的例子，作為實現(xiàn)自動駕駛的一個步驟。在“自動巡視中的Millemiglia”[4]中，汽車配備有通過滑輪附接到方向盤的直流電動機，滑輪可以根據(jù)裝有通過人造視覺接收傳感器輸入的分析控制系統(tǒng)的車載計算機的命令而移動。在Autopia計劃[5]中，，兩輛Citro?nBerlingo貨車已經(jīng)配備了自動駕駛。這里，由于常規(guī)車輛的輔助機構(gòu)由液壓系統(tǒng)供電，所以方向盤也被修改成由直流電機驅(qū)動，該齒輪連接到轉(zhuǎn)向桿上。在這種情況下，傳感器的輸入由全球?qū)Ш叫l(wèi)星系統(tǒng)提供。較新的車輛配備有電動助力轉(zhuǎn)向（EPS），其中轉(zhuǎn)向由通過小齒輪直接作用在齒條上的電動機輔助。電機轉(zhuǎn)矩取決于駕駛員所需的工作量。這種動力轉(zhuǎn)向器為我們的目的的優(yōu)點是不需要添加外部執(zhí)行器，我們可以從我們的機載計算機輕松管理輔助馬達。豐田汽車自動停車系統(tǒng)的一些發(fā)展已經(jīng)采取了這種方法[6]。本文的目的是介紹在雪鐵龍C3Pluriel車輛中自動駕駛的基于模糊邏輯的EPS控制系統(tǒng)（圖1）。這輛車已經(jīng)在私人電路上進行了測試，并且已經(jīng)實現(xiàn)了類似人的行為。圖1、雪鐵龍C3Pluriel試車。2板載設(shè)備電動助力轉(zhuǎn)向主要由轉(zhuǎn)矩傳感器和電動機執(zhí)行機構(gòu)組成。傳感器連接到轉(zhuǎn)向柱，并測量駕駛員在方向盤移動時施加的扭矩。該扭矩信號被傳送到控制/電源卡，該控制/電源卡將放大的比例功率信號發(fā)送到與轉(zhuǎn)向齒條桿接合的直流電動機。實現(xiàn)自動轉(zhuǎn)向控制的第一步是從我們在汽車上安裝的計算機管理車輪。這種自動化的方法是繞過傳感器和控制/電源卡設(shè)備，并將電源信號直接發(fā)送到電機。我們的機載計算機運行一個基于模糊邏輯的控制系統(tǒng)，產(chǎn)生一個控制模擬信號。增加了一個外部電源驅(qū)動器來取代原來的C3的電源卡。它用作計算機產(chǎn)生的模擬信號作為輸入，輸出是提供輔助電機的電源信號。車輛儀表用傳感器設(shè)備完成：一個載波相位差的GPS接收器，可以產(chǎn)生精確的定位位置和一個CAN總線接口。現(xiàn)在我們可以繼續(xù)描述開發(fā)的控制系統(tǒng)來管理直流電動機，從而控制車輛的轉(zhuǎn)向。3轉(zhuǎn)向控制系統(tǒng)已經(jīng)為轉(zhuǎn)向控制定義了兩層模糊控制器。高級層計算方向盤的目標(biāo)位置，使車輛適應(yīng)停車路線。低水平層產(chǎn)生必須由EPS輔助馬達施加以使人以類似方式移動方向盤的最佳扭矩。還有一個由待跟蹤路線的最具代表性的GPS航點的集合表示的計算參考軌跡表示[1]。3.1轉(zhuǎn)向位置控制器兩個變量用作模糊轉(zhuǎn)向位置控制系統(tǒng)的輸入，即橫向和角度誤差。我們將橫向誤差定義為車輛前方與參考軌跡段之間的距離，沿垂直于該參考段的線測量。類似地，角度誤差被描述為由軌跡的參考段和汽車的導(dǎo)向矢量形成的角度。我們定義兩個模糊變量，也稱為角度誤差和橫向誤差，每個模糊變量都有兩個左右的語言標(biāo)簽，指示車輛相對于參考段的位置。兩個變量對于每個標(biāo)簽具有一個關(guān)聯(lián)的隸屬函數(shù)，由它們的頂點定義，如圖1所示。圖2a和2b。圖2.輸入變量成員函數(shù)：a）和b）轉(zhuǎn)向位置控制器;c），d）和e）轉(zhuǎn)向扭矩控制器。系統(tǒng)的輸出是目標(biāo)轉(zhuǎn)向角，方向盤必須移動以糾正由輸入變量指示的軌跡偏差。只有一個模糊輸出變量，名為Steering，具有兩個稱為左和右的語言標(biāo)簽，其隸屬函數(shù)由單例定義。從下面的輸入數(shù)據(jù)生成轉(zhuǎn)向轉(zhuǎn)向角的規(guī)則對于彎道和直道行駛以及模糊控制和人駕駛都是相同的。人類驅(qū)動因素（規(guī)則）的定性行為在這兩種情況下是相同的，只有量化部分不同，這是通過變量模糊化在模糊控制中定義的：R1.1：如果Lateral_Error左轉(zhuǎn)向右轉(zhuǎn)R1.2：如果Lateral_Error右轉(zhuǎn)向左轉(zhuǎn)R1.3：如果Angular_Error左轉(zhuǎn)向右轉(zhuǎn)R1.4：如果Angular_Error右轉(zhuǎn)向左轉(zhuǎn)斜體中的單詞是模糊變量，其中左邊的項是輸入變量，右邊的變量是輸出變量。正常類型的詞是與每個模糊變量相關(guān)聯(lián)的語言標(biāo)簽。3.2轉(zhuǎn)向力矩控制器在這種情況下，需要三個輸入變量來控制施加到轉(zhuǎn)向輪的轉(zhuǎn)矩。第一個是方向盤的角位置誤差，即由高級模糊控制器產(chǎn)生的目標(biāo)位置與實際位置