Distributed Diagnosis for Petri Nets models with unobservable interactions via common places

In this paper we consider the case of a large plant comprising different local sites. At each site, a local diagnoser must provide the diagnosis of the site based on the local plant model (a Petri Net), the local observation and the information exchanged with its neighbors. The communication between the local diagnosers is not event-driven and the interactions between the local sites (modeled as common places) are considered unobservable (tokens can enter and exit unobservably the local Petri Net models). For this general setting we present in this paper an algorithm that allows the local diagnosers to recover completely the results of a centralized diagnoser after the completion of a communication protocol.