A stereo vision system for real time obstacle avoidance in unknown environment

Abstract

A stereo vision system for detecting and avoiding obstacles in real time in an unknown environment is described. This is only the low level of the control system architecture of a mobile robot based on the Brook's subsumption architecture. This level is based on a stereo vision algorithm that uses precomputed measurement of the ground floor disparity, and online acquired grey-level stereo images. The floor is supposed to be planar. The algorithm has been successfully implemented on a mobile vehicle, that is capable of moving around in an unpredictable environment, like a laboratory, without collision.<>