资源描述：
《6 Robust mapping and localization in indoor environment using sonar data (1)》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、RobustMappingandLocalizationinIndoorEnvironmentsusingSonarData¤JuanD.Tard´os,Jos´eNeiraDept.Inform´aticaeIngenier´ıadeSistemas,UniversidaddeZaragozaMar´ıadeLuna3,E-50018Zaragoza,Spaintardos@posta.unizar.es,jneira@posta.unizar.esPaulM.Newman,JohnJ.Leonar
2、dMITDept.ofOceanEngineering77MassachusettsAv.,Cambridge,MA02139-4307USApnewman@mit.edu,jleonard@mit.eduJune24,2002AbstractInthispaperwedescribeanewtechniqueforthecreationoffeature-basedstochasticmapsusingstandardPolaroidsonarsensors.Thefun-damentalcontr
3、ibutionsofourproposalare:(1)aperceptualgroupingprocessthatpermitstherobustidentificationandlocalizationofenviron-mentalfeatures,suchasstraightsegmentsandcorners,fromthesparseandnoisysonardata;(2)amapjoiningtechniquethatallowsthesystemtobuildasequenceofin
4、dependentlimited-sizestochasticmapsandjointheminagloballyconsistentway;(3)arobustmechanismtodeterminewhichfeaturesinastochasticmapcorrespondtothesameenvironmentfeature,allowingthesystemtoupdatethestochasticmapaccordingly,andperformtaskssuchasrevisitinga
5、ndloopclosing.Wedemonstratethepracticalityofthisapproachbybuildingageometricmapofamediumsize,realindoorenvironment,withseveralpeoplemovingaroundtherobot.Mapsbuiltfromlaserdataforthesameexperimentareprovidedforcomparison.KeywordsMapBuilding,LocalMaps,Dat
6、aAssociation,SonarSensors,HoughTransform¤ToappearinTheInternationalJournalofRoboticsResearch,200211IntroductionTheproblemofconcurrentmappingandlocalization(CML)foranautonomousmobilerobotisstatedasfollows:startingfromaninitialposition,amobilerobottravels
7、throughasequenceofpositionsandobtainsasetofsensormea-surementsateachposition.Thegoalisforthemobilerobottoprocessthesensordatatoproduceanestimateofitspositionwhileconcurrentlybuildingamapoftheenvironment.Thedifficultyofthisproblemliesinthefactthat,toobtain
8、agoodmap,anaccurateestimationofthevehicletrajectoryisrequired,butreducingtheunboundedgrowingodometryerrorsrequirestoassociatesen-sormeasurementswithaprecisemap.TheproblemofCML,alsoreferredtoasSLAM(simultaneouslocalizationandmapbu
