Visual navigation for a mobile robot: Building a map of the occupied space from sparse 3-D stereo data

This paper is concerned with enviromnent geometric modeling for a mobile robot, and more precisely deals with the problem of building a map of the occupied space from 3-D data obtained by stereovision. A general method is proposed, which produces a planar description of the occupied space consisting of a set of non-convex polygonal hulls enclosing the ground-projected 3-D features. A specific definition of the so-called "L-convex" hulls, and an efficient computation algorithm are given. Experimental results are shown at the end of the paper.