Diagnostic décentralisé des systèmes à événements discrets

We consider the detection and identification of unobservable events in the behavior of discrete event systems. We study decentralized diagnosis architectures where several sites observe the system behavior, each site having its own set of observable events. These sites do not communicate among each other but they send their local diagnostic decisions to a coordinating site that implements a simple memoryless Boolean fusion rule. Several new results are presented in the context of such architectures, with special focus on the use of conditional decisions by the local sites.