This two-year project will extend Dr. Kokar's concept-learning system, COPER, to the domain of dynamic-system identification and control. It will focus on model identification in domains where processes change their goals and characteristics at random intervals, necessitating continuous validation and adaptation of the plant model. COPER will infer a set of local models, each adequate to maintain the process in a safe operating region during its appropriate control regime. This should contribute to the theory of intelligent control within switching environments.
