Generalizing diagnosability definition and checking for open systems: a Game structure approach
Abstract
Diagnosability is the property of a partially observable system with a given set of possible faults, that these faults can be detected with certainty with a finite observation. Usually, the definition and the verification methods of diagnosability ignore the nature of the system events, controllable (by the system) or uncontrollable. In this paper we show the influence of controllability of events on the diagnosability definition and verification. We show that the classical diagnosability is a special case where we consider the whole system as controllable. Using Game Structure we generalize the definition of diagnosability by the mean of strategies. Then, Alternating-time Temporal Logic is used in order to model check diagnosability in the case of uncontrollable events. We show how the framework is suitable for one system and also for a set of interacting systems.
Origin : Files produced by the author(s)