Obstacle avoidance perception processing for the autonomous land vehicle

A description is given of an obstacle avoidance perception sensing system developed for use on Martin Marietta's autonomous land vehicle. The focus is on range-image-based methods developed to run on a high-speed systolic array processor called the Warp machine. Techniques are presented for fusing video/range sensor data, locating obstacle regions in range imagery, and mapping of the algorithm onto Warp machine. The results of applying the perception sensing system in an experimental test run are presented, and limitations of its use are discussed.<<ETX>>