The Community for Technology Leaders
Green Image
Parallel architectures have not yet achieved the generality of sequential processing, especially in real-time applications. Real-time modeling and simulation of complex systems in areas such as robotics require thousands of computations in a fraction of a second, which can be prohibitive in terms of computer hardware. This often has resulted in an emphasis on tailor-made architectures, presenting a major obstacle in applications where the architecture must host a variety of algorithms. The author proposes using a parallel processing framework to solve the real-time robot simulation problem. Robot simulation facilities are essential for testing different robotics algorithms without the costs and hazards associated with experimental prototypes. The author also proposes using small-sized, coarse-grain networks of parallel processors to execute the algorithms. These networks consist of general-purpose T800 transputer chips. The author begins by explaining the motivation for this work and then carefully formulating the problem. He addresses architectural considerations and describes robot-arm-dynamics simulation. He then focuses on parallelism in robot simulation, describing the two computational modules: the computation of the dynamics and the solution of a linear system of equations. Finally, he presents the results of implementing the framework. These results reveal great potential, in terms of cost savings and performance, for using parallel processing in robotics computations.
MIMD, parallel processing, real-time computing, robotics, simulation

A. Y. Zomaya, "Parallel Processing for Real-Time Simulation: A Case Study," in IEEE Concurrency (out of print), vol. 4, no. , pp. 49-62, 1996.
82 ms
(Ver 3.3 (11022016))