冗長系逆運動学マップ計算の並列分散アーキテクチャ

Abstract

A high-speed computation algorithm and a developed hard-ward architecture are described in this paper which are used for solving the inverse kinematics method of a redundant robot by a nonlinear mapping method. The nonlinear mapping algorithm computes the nonlinear map between fingertip positions and joint angles of the robot by a diffusion-based parallel dispersed method. However, to compute the high density nonlinear map by this algorithm, there are two problems, that is enormous computing time and many processors are required. In this paper, we propose one way of computing high density map with the reasonable number of processors and computing time by the divided parallel dispersed method. Also, we have designed a hard-ware architecture and composed parallel CPU computer system and then made some experiments to evaluate the performance of this system.