Parallel computing of symbolic robot models and control laws: Theory and application on advanced multiprocessor architectures

Abstract

This paper is devoted to advanced parallel computer architectures based on transputers and array processors. The goal is to achieve the computation of various robot control structures in a minimum time. Since the most of the advanced position/force control laws employ robot dynamic compensation, the attention will be paid to the computation of robot dynamics on parallel architectures. Starting from the most efficient customized symbolic robot equations based on Newton-Euler's formalism, the scheduling algorithms for the serial, parallel and the array processor architectures are developed. The algorithms take into account both the load balancing and the communication overhead during execution of scheduled tasks. Experimental results on transputer based pipeline and parallel architectures and on an array processor are presented.