Knowledge-Based Real-Time Control

Abstract This paper describes some experiences in using knowledge-based systems in real-time control. We distinguish two types of expert control. The first, direct expert control, uses the knowledge-based system directly to influence the process through a control signal. The second approach, supervisory expert control, influences the process in an indirect way. A real-time expert system shell is described, using progressive reasoning. This approach presents a combination of direct and supervisory control. The direct expert control approach has been investigated by using the phase plane and has proved to be useful for many processes. Supervisory expert control experiments were carried out using heuristic adaptation and pattern recognition. Finally a supervisor mechanism for an adaptive system is described.