A Parallel Newton-Euler Formulation for Fast Dynamic Simulation of Robot Manipulators

Advanced control strategies require the inclusion of the dynamical model of the robot arm in the control law. However, the dynamics consist of a highly coupled and non-linear set of equations. Thus, this complexity has always presented a major obstacle in real-time dynamic control applications. The computationally efficient solution of this problem will lead to a better comprehension of the key factors affecting robot operations. This work describes a solution of this problem by employing a parallel processing approach. The dynamics are computed by using a semi-customised Newton-Euler formulation. The algorithm is distributed over a highly-coupled multiple-instruction multiple-data stream (MIMD) computer architecture. The computer system is constructed from general purpose (VLSI) building blocks called (TRANSPUTER. The cost-effectiveness and speed of the scheme is demonstrated by a case study (PUMA 560 robot arm). The communication issues between the different processors are discussed. Speed-up results are included to show the superiority and advantages of the parallel approach.