Approximation Algorithms, Sensor Selections and Information Sharing

This paper discusses recent results related to the approximation of optimal sensor selection and information sharing policies for the supervisory control of discrete-event systems. Although the framework of supervisory control is used, the results and methods contained herein are also applicable to fault diagnosis systems. It is shown how the problems of computing an optimal sensor selection is related to the problem of computing edge-colored direct graph st-cuts. Heuristic approaches to the approximation of optimal solutions to the sensor selection problem are proposed based on graph cutting and integer programming methods. It discussed how the graph cutting conversion method can be used for other problems of partial information in supervisory control such as a communication selection problem for decentralized control systems.