Optimal robot decomposition for fast path planning

Abstract

The paper presents a method for path planning by decomposition and parallel search. The manipulator is decomposed into chains of consecutive links and joints. A set of postures is defined to represent the motions of each chain. A collision table combines these sets and identifies collisions of the robot. A two phase online planning method constructs a path for each chain independently, and then synthesizes them into a collision free robot path. A backtracking scheme is introduced to improve the success rate, and a prescription is given for optimal decomposition of a robot for reduced planning time and improved success rate. An example of a 7-DOF robot is given to demonstrate the method, and to show its short online planning time.