REPORT 9606 -- MASc Thesis
Software for State-Event Observation Theory and its Application to Supervisory Control

Yiqi Zhang
Abstract

Previous to the author's work, in an attempt to create a more powerful model reduction technique for DES control design, an observation theory called LW theory based on both event and state information has been developed at the University of Toronto. The author's work is to design software for this theory and then try to apply the theory in the RW control framework.

A comprehensive computer program for LW state-event observation theory is developed. Various procedures defined in LW theory such as Project, Refine and Quotient can be computed by the software. The general design and choice of data representation are discussed extensively. The key algorithms including RCP are carefully studied. Several examples including an industrial one are used to illustrate the various applications of the software.

A new architecture for applying the LW observation theory in RW control framework is presented. The model representation is carefully studied. The traditional supervisor design is extended to a new supervisor design procedure which can handle the nondeterministic situation with possible silent event. Finally, the property of control consistency is also guaranteed.