Practical parallel processing for real-time adaptive control of mechanical arms

This article is concerned with efficient and fast computations of adaptive control algorithms for robot manipulators. Adaptive controllers employing the highly coupled and nonlinear dynamics of manipulators are computationally complex, which presents a major obstacle in their real-time implementation for industrial applications. A solution to this problem is suggested by utilising a parallel processing approach. The controller algorithm is implemented for a PUMA 560 robot manipulator with six degrees of freedom utilising INMOS T800 transputer network, for which programming is performed in the ANSI C language. The execution times achieved by the distributed algorithm are well within the limit acceptable for real time control