3D Robotic Mapping with Probabilistic Occupancy Grids

We propose a new approach to 3D environment mapping from a mobile robot, using visual information provided by alow-cost stereo vision system. Depth information recovered from disparity maps serves as input for our method that uses probabilistic modelling to determine if a region is occupied or freein an occupancy grid approach. There is no similar approach in the literature, based on statistical modellingof the errors from stereo measurements and robot’s motion, so this is the main contribution of this work. The resulting map canbe directly used in tasks like obstacle avoidance, navigation, localization, exploration and path planning,enabling a mobile robot to perform autonomous navigation in its environment. Real experiments done with a Pioneer 3-AT mobile robot equipped with a Minoru stereowebcam show the efficiency of our mapping approach