Adaptive control with expert system based supervisory functions

A supervisory architecture is presented that makes use of a time distributed interactive diagnostic paradigm to detect and treat problems with adaptive controllers. Characteristics of the diagnostic process for problems with adaptive controllers are discussed from a control engineer's view point, and it is shown that, in general, diagnosis evolves over time as information becomes available through experimentation with the adaptive controller. An expert system framework is described which supports this style of diagnosis and a detailed treatment of the knowledge representation used is given. Simulations of the expert supervised adaptive controller are provided, and it is shown that even in the case where event driven diagnostic cannot be used because of incomplete sensory information, the interactive supervision paradigm is able to resolve performance problems^ in an adaptive controller