Expert aided adaptive control

Abstract

The idea of a two-level control scheme, the A/E (adaptive/expert) controller, is introduced as a way to optimize the behavior of an adaptive controller using expert system techniques. The proposed structure uses a constant expert system sampling interval and 14 features to describe the state of the adaptive controller. Key issues discussed are the transformation of signal domain data for use by the expert system module of the A/E controller, and the use of a three-part knowledge base to formulate corrective actions of the expert system module. Simulation results show a prototype of the A/E controller in comparison to a standard adaptive controller. These results indicate that even a simple prototype of the A/E controller can improve the performance of an adaptive controller over that found when using conventional adaptive techniques. Preliminary results indicate that the A/E controller will result in significant improvements in adaptive control reliability and performance.<>