The Control of Real-Time Discrete-Event Systems Subject to Predicate-Based Constraints

Ph.D. Thesis by Tan-Jan HO, 1997


Abstract

A unique systematic design methodology is developed for achieving efficiency in solving some useful control problems of real-time discrete-event systems. To achieve this, an extended modelling formalism, which allows the incorporation of data variables, is proposed for describing real-time discrete-event systems. Based on this formalism, a new state-feedback control framework which can exercise two types of control mechanisms: enabling/disabling and forcing is established. Two important control problems are handled in this thesis. The first problem involves finding solutions which ensure that predicate-based constraints are satisfied invariantly in a closed-loop real-time discrete-event system. To tackle this problem leads to the development of a self-contained theory based on an innovative analysis of event trajectories of a system and a strategy for manipulating control mechanisms. A novel decision algorithm, employing a hybrid dynamic programming/constrained-search method, offers a baseline for handling forcing and thereby plays a key role in the derivation of sufficient conditions which guarantee this control problem is solvable. Furthermore, a procedure for controller synthesis is obtained based on a crucial connection between solvability conditions of this problem in this theory and an adapted notion of control-invariance in Ramadge-Wonham theory. The second control problem is how to find solutions which ensure that the first control problem is solvable and deadlock-freedom is achievable. Sufficient conditions are presented for the existence of solutions to this problem. For both control problems, synthesis procedures in pseudocode are provided. As application examples, some control problems are examined: a traffic-light in a subway system, a transfer line in a manufacturing system, and two rods in a nuclear power plant. The feasibility of this systematic design approach is effectively demonstrated by these examples in terms of computational effort and explicit representations of control policies. As a whole, this thesis combines mathematical theory and quasi real-world applications.