Distributed VLSI Architectures for Fast Jacobian and Inverse Jacobian Formulations

摘要

The rapid development in VLSI technology makes it possible to implement highly complicated and time consuming algorithms to suit real-time applications. Parallel processing techniques can now be used to reduce computational time for models of robot manipulators. The development system used to implement the algorithms consists of an INMOS TRANSPUTER ) a VSLI single-chip computer) running the OCCAM concurrent programming language. This system is used to construct and evaluate the performance and cost effectiveness of several proposed methods to solve for the JACOBIAN and INVERSE JACOBIAN problems with special attention to the case of the robot operating in the neighbourhood of singular points. Detailed analysis is performed and successful results are obtained for a 6 dof robot arm (PUMA 560). Execution time comparisons between Von Neumann (uniprocessing) and parallel processing architectures are also included to show the superiority of the latter approaches.