EKF Localization andMappingbyUsingConsistent SonarFeature withGivenMinimumLandmarks
Sejin Lee, Jong‐Hwan Lim · 2006
TheSLAMorlocalization needs successful data association ofthedetected feature withlandmarks. Well described features oftheenvironment areessential forgooddata association. Inthis paper, thelocalization oftherobot isexecuted bytheextended Kalmanfilter (EKF)withgiven minimumlandmarks oftheenvironment. Consistent features forlocalization areextracted byusing onlysparse sonardata. Features areextracted byusing asonar data clustering fromafootprint-associ ation (FPA)methodanda feature fitting fromaleast squares (LS)method to overcome challenges associated withsonar sensors, suchasawidebeamaperture andaspecular reflection effect. The extracted features are, also, evaluated asapost-processing through theprobabilistic association whichassociates the extracted feature withtheweighted average probability ofthegrids that arelocated within theareaofposition uncertainty ofthefeature. Theproposed methods havebeentested inareal homeenvironment with amobile robot.