SLAM with Expectation Maximization for Moveable Object Tracking.pdf

《SLAM with Expectation Maximization for Moveable Object Tracking.pdf》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、The2010IEEE/RSJInternationalConferenceonIntelligentRobotsandSystemsOctober18-22,2010,Taipei,TaiwanSLAMwithExpectationMaximizationforMoveableObjectTrackingJohnG.RogersIII,AlexanderJ.B.Trevor,CarlosNieto-Granda,andHenrikI.ChristensenAbstract—Thegoalofsimultaneouslocalizationandmappingvalidatedinan

2、indoorenvironment.Thispaperisorganized(SLAM)istocomputetheposteriordistributionoverlandmarkasfollows:AfteradiscussionofrelatedworksinsectionII,poses.Typically,thisismadepossiblethroughthestaticworldweexplainouralgorithmforhandlingmoveablelandmarksassumption–thelandmarksremaininthesamelocationinS

3、LAMbyfirstdiscussingtheSmoothingandMappingthroughoutthemappingprocedure.Somepriorworkhasaddressedthisassumptionbysplittingmapsintostaticand(SAM)algorithmwithourextensionsinsectionIII.Thedynamicsets,orbyrecognizingmovinglandmarksandtrackingexperimentalprocedurewillbedescribedinsectionIVandthem.Inc

4、ontrasttopreviouswork,weapplyanExpectationexperimentalresultswillbeanalyzedinsectionsVandVI.MaximizationtechniquetoagraphbasedSLAMapproachFuturedirectionsforthisworkwillbeoutlinedinsectionandallowlandmarkstobedynamic.ThebatchnatureofVII.thisoperationenablesustodetectmoveablelandmarksandfactorthe

5、moutofthemap.WedemonstratetheperformanceofthisalgorithmwithaseriesofexperimentswithmoveableII.RELATEDWORKlandmarksinastructuredenvironment.ThesummarypapersbyDurrant-WhyteandBailey[8],I.INTRODUCTION[2]serveasanexcellentintroductiontotheSLAMproblem.TherearetwobasicstrategiesforsolvingtheSLAMprob-R

6、obotsneedthecapabilityoflocalizationandmappinglem.ThefirstapproachusedbytheroboticscommunitywastoperforminmanyapplicationdomainssuchasmobilebasedupontheuseoftheExtendedKalmanFilter(EKF)formanipulation.Anessentialpartofnavigationistheabilitytomobilerobotlocalization.TheEKFwasusedby[5]and[6]remainl

7、ocalizedintheirenvironment.IntheSimultaneousformobilerobotlocalizationusinganapriorimap.TheLocalizationandMapping(SLAM)problem,weareinter-firstviableSLAMalgorithmbasedontheEKFwasreportedestedinestimatingthemostlikelypositions