Implementation of Dynamic Obstacle Avoidance on the CMU NavLab

Abstract

The design and implementation of local obstacle avoidance algorithms on the CMU NavLab testbed for autonomous navigation are presented. The CMU NavLab carries various computing systems, including three Suns, one Warp systolic array processor, and several Intel 386 real-time processors, for the purpose of processing sensor information and generating vehicle motion commands. The obstacle avoidance algorithm presented is part of a reflexive path tracking scheme. To handle such situations where obstacles are present on the predefined path, an obstacle avoidance algorithm is added to the reflexive path tracking scheme whereby when an obstacle is detected on the predefined path, the vehicle control is transferred to the obstacle avoidance algorithm which guides the vehicle around the obstacle. This combination results in a scheme that reacts to the changing and uncertain environment. A subgoal selection algorithm and a steering control algorithm are also discussed.