An approach to implementing a knowledge-based controller

The results of building a knowledge-based adaptive controller for a DC-motor-driven, angular-positioning servo-trainer are presented. The approach used is general in the sense that tools have been built which will support knowledge-based identification, control law calculation, and controller implementation in response to changes in the system plant. However, the implementation is restricted to lead-lag compensation of a single-input, single-output linear, continuous plant and is currently limited by the accuracy of the model identification results. The system is capable of executing the cycle of identification, controller design, and implementation in approximately 10 s for a servo trainer used for classroom laboratory experiments. A significant analytical shortfall exists in that an assessment of the stability of such a system is not possible using currently available techniques. The system is described, and the tools developed for an inference engine to support reasoning about time-sensitive events are discussed