资源描述：
《Autonomous Vehicle Navigation in Rural Environments Without Detailed Prior Maps》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。
1、2018IEEEInternationalConferenceonRoboticsandAutomation(ICRA)May21-25,2018,Brisbane,AustraliaAutonomousVehicleNavigationinRuralEnvironmentswithoutDetailedPriorMapsTeddyOrt1,LiamPaull1,2,DanielaRus1Abstract—State-of-the-artautonomousdrivingsystemsrelyheavilyonde
2、tailedandhighlyaccuratepriormaps.However,outsideofsmallurbanareas,itisverychallengingtobuild,store,andtransmitdetailedmapssincethespatialscalesaresolarge.Furthermore,maintainingdetailedmapsoflargeruralareascanbeimpracticableduetotherapidrateatwhichtheseenviron
3、mentscanchange.Thisisasignificantlimitationforthewidespreadapplicabilityofautonomousdrivingtechnology,whichhasthepotentialforanincrediblypositivesocietalimpact.Inthispaper,weaddresstheproblemofautonomousnavigationinruralenvironmentsthroughanovelmaplessdrivingfr
4、ameworkthatcombinessparsetopologicalmapsforglobalnavigationwithasensor-basedperceptionsystemforlocalnavigation.First,alocalnavigationgoalwithinthesensorviewofthevehicleischosenasawaypointleadingtowardstheglobalgoal.Next,thelocalperceptionsystemgeneratesafeasib
5、letrajectoryinthevehicleframetoreachthewaypointwhileabidingbytherulesoftheroadforthesegmentbeingtraversed.Thesetrajectoriesareupdatedtoremaininthelocalframeusingthevehicle’sodometryandtheassociateduncertaintybasedontheleast-squaresresidualandarecursivefiltering
6、approach,whichallowsthevehicletonavigateroadnetworksreliably,andathighspeed,withoutdetailedpriormaps.Wedemonstratetheperformanceofthesystemonafull-scaleautonomousvehiclenavigatinginachallengingruralenvironmentandbenchmarkthesystemonalargeamountofcollecteddata.
7、I.INTRODUCTIONAutonomousdrivinghasthepotentialtodrasticallyim-proveourlives.Todate,thevastmajorityoffieldedau-tonomousvehiclesfocusononeoftwoscenarios:Fig.1:MaplessNavigationusingSparseTopologicalMaps.1)Lanefollowingonwell-markedstructuredhighwaysTop:Thecrowd-s
8、ourcedtopologicalmapfromopenstreetmap.orgisshownasredsegmentsconnectingyellowvertices.Thepoint2)Urbannavigationbasedonextremelypreciseandman-cloudobtainedfromaVelodyneHDL-64lasersc
