Supervisory control of dense real-time discrete-event systems with partial observation
Ahmed Khoumsi · 2003
In supervisory control theory, the basic task of the supervisor is to disable certain events of the plant so that the obtained behaviour lies within a given specification. We propose a method which extends this theory with the following two points. First, the plant and the specification contain temporal constraints and are described by a model called Timed Automata (TA). Second, the supervisor has only a partial observation of the behaviour of the plant. The problem that arises with the TA model is that the state space can be infinite. Recently, we proposed a method to finitely represent the state space which generates less states than the well-known region graph approach. Its principle consists of transforming a TA into a Finite State Automaton (FSA) using two special types of events: Set and Exp. Such a FSA is denoted se-FSA. In this article, we propose a method for the supervisory control of timed discrete event systems that are modelled by TA and partially observable. We use the above-mentioned transformation procedure for representing the plant and the specification by two se-FSAs. Then, we develop a procedure for generating the supervisor from the two se-FSAs that represent the plant and the specification. We also propose a supervisory control architecture.