Distributed diagnosis using predetermined synchronization strategies in the presence of communication constraints

We consider distributed fault diagnosis in a discrete event system modeled as a nondeterministic finite automaton that is observed at multiple observation sites through distinct natural projection maps. The majority of previous work in this setting has focused on local observers (diagnosers) that are separately implemented at each observation site, i.e., these diagnosers do not attempt at any point to refine their diagnostic information by exchanging information among themselves, and typically rely exclusively on communicating their decision (fault or no fault) to a coordinator that is responsible for making the ultimate diagnosis decision (e.g., co-diagnosability). In our previous work, we extended these techniques to allow local observers/diagnosers to communicate their state estimates along with diagnostic information (normal or fault condition) to a coordinator that fuses local diagnostic information and subsequently provides this refined information to the different observation sites (to aid them in taking immediate or future diagnostic decisions). In this work, we extend these synchronization strategies to cases where a coordinator is not present and there may exist communication constraints between the exchange of information among local sites (i.e., each local diagnoser may only exchange information with a subset of the local diagnosers). We verify diagnosability in this setting via the construction of an appropriate composition of local diagnosers, which is able to capture the refinement of information under the given synchronization strategy and communication constraints.