Computational architecture for linear n-processor array implementations of composite rigid-body spatial inertial matrix algorithms

摘要

A computational architecture is proposed which meets the requirements for real-time calculation of the composite rigid-body spatial inertial matrix using a linear n processor array. This architecture provides for expansion to larger linear-array structures and can be modified to act as the basic structure for a 2D computational array. The multiple communication paths and internal parallelism provided by the C40 processor make this architecture amenable for use in computing architecture dependent O(n/sup 3/) algorithms with few modifications. This assumption means that the proposed architecture can be used as the central computing node for solving the entire forward dynamics problem in the control of robotic manipulators. The use of the TMS320C40 digital signal processor as the primary processing element reduces many of the hardware and timing problems encountered with other types of processors.