Efficient Abstractions in Hierarchical Supervisory Control of Discrete-event Systems

In this thesis we study two problems in hierarchical supervisory control of discrete-event systems. The first problem is controller reduction, which is to find smaller abstractions of hierarchical controllers for easier understanding. The second problem is controller design, which is to find smaller abstractions of systems for high-level controller design. We employ natural projections to generate abstractions, thus incorporate the two problems into one optimization problem in which we try to minimize the size of the observable event set of the abstractions, subject to the constraints that the abstractions still achieve maximally permissive and nonblocking control. An optimal solution is proposed for the controller reduction problem, and a suboptimal solution is proposed for the controller design problem.MAS