Localization and Decomposition in Supervisory Control of Discrete-Event Systems

Supervisor localization procedure can be used to construct the local controllers relevant to each component agent in a discrete-event system. This procedure is based on state reduction of monolithic supervisor with respect to each set of controllable events corresponding to each component agent. In this paper, observation and coobservation properties which preserve through supervisor reduction and supervisor localization procedures are investigated. It is proved that the supervisor reduction procedure preserves observation properties of a monolithic supervisor by self loop transitions at some states in the reduced supervisor, as well as the supervisor localization procedure preserves coobservation properties. A new method based on decomposition of the reduced state supervisor is proposed to localize a monolithic supervisor and to construct the local controllers such that the observation properties from the monolithic to distributed supervisory control are preserved. An example is illustrated to verify the effectiveness of the proposed technique.