Location andOrientation Detection ofMobile Robots Using SoundField Features under Complex Environments

Inthis paper, thefeasibility ofutilizing soundfor robot's posedetection isinvestigated, anda novelandrobust robotlocation andorientation detection methodbasedonsound fieldfeatures fornoisyenvironment isproposed. Unlike traditional methods, theproposed methoddoesnotexplicitly consider thecharacteristic ofdirect pathfromsoundsourceto microphones, norattempt tosuppress theeffect ofreverberations andnoise signals. Instead, itutilizes thesoundfield features ofa robotat different location and orientation in a normal environment. Thesoundfield feature iscaptured byusinga probability distribution estimation methodcalledGaussian Mixture Model(GMM).Theexperimental results showthatthis methodcandetect robot's location andorientation underboth line-of-sight and non-line-of-sight conditions usingonlytwo microphones andisrobust toenvironmental noise. Moreover, it canalsosolve themicrophones' mismatch problem andcanbe applied tobothnear-field andfar-field conditions. Sincethis methodcanprovide global location andorientation detection, it issuitable tofusewithotherlocalization methodstoprovide initial conditions forreduction ofthesearch effort, orprovide the compensation forlocalizing certain locations thatcannotbe detected using other localization methods. IndexTerms- Robotlocalization, Robot'sorientation detection, GMM.