Knowledge-Based Real-Time Control: A Parallel Processing Perspective

Knowledge-based real-time control problems can be usefully viewed as dynamic resource allocation problems. Analysis of various real-time applications and real-time AI models reveals that real-time control problems require the problem solving capability of knowledge intensive methods coupled with the control mechanisms of operating systems. Moreover, there is an opportunity and need to exploit parallelism inherent in real-time control problems. We describe a user-programmable concurrent computation model which blends the capabilities of knowledge-based systems and operating systems. We also propose a novel set of performance measures useful for real-time AI systems.