DETECTIONOFDRIVABLECORRIDORSFOR OFF-ROADAUTONOMOUS NAVIGATION
Gary R. Bradski · 2006
Thispaper describes ahierarchical Bayesian network usedforsegmenting desert images anddetecting offroaddrivable corridors for autonomous navigation. Unlike theembedded hidden Markov model theBayesian network presented inthis paper cansuccessfully accountfornatural dependencies between neighboring pixels inboth imagedimensions making itmoresuitable foralarger class ofimages. Themethod described here wasdeveloped within theStanford racing teamthat wontheDARPAGrandChallenge 2005after drivingover130miles autonomously intheNevada desert. IndexTerms-Mobile robot motion-planning, imagesegmentation, hidden Markovmodels