Supervisory Control of Discrete Event Systems with State-dependent Observability

This paper considers a new class of discrete event systems under partial observations. The problem is presented within the background of a manufacturing process where workpieces are loaded and transported, and this process is controlled with the partial information collected by sensors. The model extracted is novel because the observation of an event does not only depend on an event itself, but also the state where the system stays. Two standard problems are discussed in this paper: supervisor existence problem and supervisor synthesis problem. With a natural revision of observable languages, a necessary and sufficient condition is given for the existence of a supervisor. For supervisor synthesis problem, two algorithms are developed: one algorithm is to check the properties of a control specification given by a regular language, and the other one is to synthesize a supervisor if the properties hold. Within the background of manufacturing systems, an example is illustrated to show how the algorithms are applied to practical computing.