| This thesis proposes a reconfiguration framework for the class of Discrete Event Systems (DES) in which the availability of observation means is subject to change over time. The designed reconfiguration yields a more tolerable controller than the close work in the literature. Starting with a maximally permissive controller that has a full observation of its DES, a mega-controller is designed to monitor the observation set of the DES controller and its state continuously. Upon a change in the observation set, the mega-controller reconfigures the controller. The reconfiguration strategies embedded in the mega-controller are created based on an important concept, "finite time observation policy" proposed in this thesis.; By applying the reconfiguration strategies, the mega-controller can reconfigure its control in order to accommodate the change of observation means availability while minimizing the out-of-control situations. The mega-controller's behaviors are modeled by Finite Automata (FA), in which, the observation means failure and repair are considered as events, and control reactions, i.e. reconfiguration, to those events are included. The model can further be converted to a Markov decision process to facilitate the optimization of the reconfiguration strategies.; This thesis also explores the application of the developed theory on the Programmable Logic Controller (PLC) programming. The configurability is integrated into the design of Ladder Logic Diagrams (LLD). The reconfiguration theory can systematically and comprehensively analyze the situations caused by the change of availability of input information, and propose a proactive measure to prevent out-of-control situations. Conversion algorithms are proposed in order to bridge the gap between LLD and FA controllers as mentioned above. Pseudo codes or mathematical models are presented for the conversion algorithm, which can be programmed or utilize/integrate with some commercial software such as simulation tools of PLC and integer programming tools. This makes it possible to apply the proposed theory to large and complex real systems. |
