Multiarm Collision Avoidance Using The Potential-Field Approach

Abstract

A general conceptualization of Khatib's potential-field approach to manipulator collision avoidance is presented. The potential-field approach is shown to consist of four algorithms: a. repulsion away from obstacles, an attraction towards a goal, a method of combining these, and a resulting method of incrementing the arm. Alternatives for these algorithms are discussed. A multiple-robot system demonstrating the concepts is presented. The system uses a detailed rigid model of the entire arms and surrounding objects to avoid collisions. The system operates in close to real time, and is demonstrated with two PUMA robots moving concurrently. Results are applicable to any type of anthropomorphic arm, including the Remote Manipulator System.