Parallel algorithms for robot dynamics

Abstract

The Department of Energy recently established a Center for Engineering Systems Advanced Research (CESAR) at the Oak Ridge National Laboratory (ORNL). The Center's charter is to conduct long-range energy-related research in intelligent control systems. This paper reports initial results in developing parallel algorithms for efficiency enhancement in real-time solutions of manipulator dynamics equations. Two approaches to the solution of the inverse dynamics problem are discussed. The first is concerned with the implementation of Newton-Euler equations in multiprocessor architecture with emphasis on asynchronous algorithms and interprocess communication. The alternative approach is based on an explicit state description of the manipulator dynamics, obtained using computer-assisted analytic simplifications of the symbolic Lagrange-Euler equations. Multicomputer and multiprocessor implementations are discussed. The construction of a compact knowledge-base in terms of associative memories is also suggested, to allow solutions of the inverse dynamics based on similarity. Future directions are also outlined. This research is an integral part of a large systems integration effort with complementary tasks in strategy planning, sensor fusion, etc.