Sampling-Based Retraction Method for Improving the Quality of Mobile Robot Path Planning

Abstract

This paper presents a method for improving the quality of the initial path produced by the probabilistic roadmap (PRM)-based mobile robot path planner. The sampling-based retraction method modifies the initial path to achieve approximate maximum safety by removing unsafe and redundant sections. The updated directions and distances of the waypoints on the initial path are determined by approximately modeling clearances around the initial paths using random samples. The proposed method can control the update speed to induce smooth convergence. The performance of the proposed method was verified by simulation.