LoopClosing inTopological Maps
Kristopher R. Beevers · 2005
Inordertocreate consistent mapsofunknown environments, a robotmustbeabletorecognize whenit hasreturned toapreviously visited place. Inthis paper, we introduce anevidential approach totheloop-closing problem fortopological maps,basedontheDempster-Shafer theory ofevidence. Inourapproach, therobotmakesahypothesis whenever itmayhaverevisited aplace. Itthenattempts to verify hypotheses bycontinuing totraverse theenvironment, gathering evidence thatsupports (orrefutes) thehypotheses. Wedescribe methods formanaging belief aboutmultiple loop- closing hypotheses, andfordetermining abelief assignment givena pieceofevidence. We alsodiscuss methodsfor reducing thefalse alarmrateofourloop-closing algorithm, andprovide simulated andreal-world experimental results thatverify theeffectiveness ofourapproach.