A Tabu Search-based Approach for Online Motion Planning

Abstract

In this paper an online motion planner is developed to govern the movements of mobile robots during their explorations. By using the Tabu search meta-heuristic method, a set of tabu (i.e., forbidden) moves are defined in each iteration of the search to confine the robot's navigable locations, and guide it toward the goal. Based on range-sensor readings and the cost function value defined for each ray, the robot is attracted to certain obstacle vertices, and moves along a path consisted of lines connecting the vertices of different obstacles. Hence, the resulting trajectory has a local shortest path quality similar to the visibility graph roadmap. The planner also takes advantage of random moves when trapped in dead-ends. Different experiments have shown the efficiency of this approach.