A Dynamic Obstacle-Avoidance Scheme for Autonomous Robots

Abstract

The real-time path-planning problem for autonomous robots is considered with the presence of arbitrary roving and static obstacles in the workspace. It is solved as a dynamic shortest route problem by adapting Dijkstra's algorithm to the workspace motion-scene graphs generated at successive scan intervals. In this obstacle-avoidance scheme, the robot takes the relative importance of detected obstacles into consideration. Simulation results show that the success rate of collision-free navigation for the new approach is 90% compared to 60% for the classical approach of wait/temporary relocate-resume. However, the total time and the distance travelled are about 10% longer for the new scheme than that obtained using the traditional technique. The numerous needs for, and benefits of self-driven robots make the research both justifiable and desirable.