Diagnosability Of Asynchronous Discrete Event Systems in Partial Order Semantics

In truly asynchronous, distributed systems, neither global state nor global time are available. Automata-based diagnosis therefore reaches its limitations there; a different approach, based on Petri net unfoldings, was proposed in [9]. It is motivated by the problem of event correlation in telecommunications network management, and uses only local states, in combination with a partial order model of time. Diagnosis is performed by correlation of the observed partial order alarm patterns and partial order executions of the system model. As in the classical automata setting, the presence of invisible transitions raises the problems of observability and diagnosability of a given system. In this paper, we give a definition of weak and strong observability and diagnosability in terms of partially ordered executions, and characterize diagnosable systems; the characterizing property can be effectively verified using a finite complete prefix of the net unfolding