Robot Inverse Dynamics on a Concurrent Computation Ensemble

摘要

Advances in VLSI technology provide a strong incentive to configure the ''brain'' of future intelligent robots as concurrent computation ensembles. The design of real-time operating systems for such machines is an important objective of the CESAR program. This paper addresses the development of ROSES, a prototype software system intended to provide expert scheduling capabilities in mapping precedence-constrained task graphs onto a computation ensemble. This problem being strongly NP-complete, ROSES attempts to achieve near-optimal solutions by combining heuristic techniques (to handle time complexity) with special instances of abstract data structures (to handle space complexity). The technique is applied to the solution of the inverse dynamics equations of a robot arm and achieves high efficiency and proper load balance. Critical concepts in the longer-term development of a Virtual Time Intelligent Machine Operating System (IMOS/VT) are also discussed. (ERA citation 10:042639)