基于概率方法的機器人定位.pdf

1、圖書分類號：UDC：工學(xué)碩士學(xué)位論文基于概率方法的機器人定位碩士研究生：導(dǎo)師：申請學(xué)位級別：學(xué)科、專業(yè)：所在單位：答辯日期：授予學(xué)位單位：雷亞莉高慶吉教授工學(xué)碩士控制理論與控制工程自動化工程學(xué)院20073東北電力大學(xué)東北電力大學(xué)碩f：學(xué)位論文AbstractThispapersystematicallyanalysesandstudiesthelocalizationapproachesforautonomousmobilerobots

2、navigationandmainlyontheprobabilitybasedlocalizationapproachesTheresearchtopicsincludeapproachesofrobotsposetracking，Markovlocalization，ParticleFilterandotherimprovedPFmethodTwodifferentgloballocalizationmethodsbasedonBa

3、yesianestimationtheoryaremainlyinvestigatedinthepaperThefirstoneistheSRL(SensorResettingLocalization)approachwhichproposedtoovercometherobotkidnappedproblemItinheritstheadvantageofPF(ParticleFilter)，representingmultipleu

4、nimodaldistributioninsteadofcalculatingprobabilitydistributionfunctionexplicitlyPFrepresentsthebeliefbyasetofsamples，whichisdrawnfromtheposteriordistribntionovertherobot‘sposeSThePOseestimationalgorithrnofPFneedslessstor

5、agespacethanpositionprobabilitygridsmethodandiswithhigheraccuracyandfastercomputingefficiencyAtthesametimesensorresettingoperationisaddedtoPFwhichsolvesthesampledeficient，SensorResettingLocalizationresamplesanumberofsamp

6、lesbasedonthesensordatawhenthemostrecentsensorreadingdoesnotagreewithwheretherobotthinksitiSalSOmeanstherobotiSlostSRLiSrobusttomodelingerrorsincludingunmodelledmovementsandsystematicerrorsItcanbemorequicklyrecoverfromki

7、dnapstatethantraditionalPFASRL(AdaptiveSensorRescUingLocalization)algorithmformobilerobotisproposedbasedonSRLItadaptivelyadjuststhenumberofnewneededsamplestOavoidingsamplesdeficiencyusesfewsamplesrealizerealtimerobotglob

8、allocalizationanddouble—updatessamplesbysensorinformationwhichguaranteestheprecisionoflocalizationsatisfiestherequirementofpracticalsystemInthepaperweparticularlydiscussthemotionmodelandsensormodelforASRLandcompareASRLto