Principles of distributed model-based diagnosis of active systems

The notion of distributed model-based diagnosis (DMBD) of a class of discrete-event systems, namely active systems, is introduced with the support of six requirements. First, the active system properties relevant to DMBD are established. Second, a variety of observers are allowed, each observer watching the system under a different view. Third, uncertain observations are considered. Fourth, the diagnostic process is expected to be performed in a distributed, virtually parallel, way. Fifth, the diagnostic process is required to be supported by some computational optimization criteria. Finally, candidate diagnoses are supposed to be given incrementally, at different run-time points. Emphasis is put on the last two requirements. In particular, an algebra is introduced and equivalence rules for algebraic expressions are defined to substantiate the optimization of the diagnostic process. Then, a method for producing an increasingly refined complete set of candidate diagnoses is envisaged.