Robot Motion Planning: A Distributed Representation Approach

Abstract
In this paper, we propose a new approach for planning the motion of robotic systems among obstacles, which is based on a distributed representation of the world model. Within this approach, we designed and implemented a general purpose path planner with five new capabilities: (1) It is able to generate very complex motions for robots with many degrees of freedom. In particular, we succeeded in generating complex paths for a 10 DOF non-serial manipulator arm made with both revolute and prismatic joints. (2) It is drastically faster (between 1 and 2 orders of magnitude) than existing systems on a sequential computer. We generated complex paths for a 3 DOF bar in a 2D workspace in about 1 second on a MIPS-based workstations, as opposed to minutes or even tens of minutes for other algorithms. The planner outputs a path for the robot in configuration space, while the goal is specified in operational space. Hence, the inverse kinematic problem is completely avoided. Furthermore, any kind of redundancy of the robot arms can be handled without modification.