Discrete approximation and supervisory control of continuous systems

Addresses the following hybrid control problem: a continuous plant (its state evolving in Euclidean space) is to be controlled via symbolic output feedback-both measurement and control signal "live" on finite sets of symbols. We adopt the following approach: the hybrid problem is first translated into a purely discrete problem by approximating the continuous plant model by a (nondeterministic) finite-state machine. By taking into account past measurement and control symbols, approximation accuracy can be improved and adjusted to the specification requirements. Supervisory control theory for discrete-event systems (DES) is then applied to find the optimal controller which enforces the specifications. As the behavior of the approximating automaton is guaranteed to contain the behavior of the underlying continuous plant model, the controller also forces the latter to obey the specifications.