Deadlock avoidance in FMS based on structural theory of Petri nets

This paper presents an efficient deadlock avoidance method for FMS combining prevention and avoidance approaches using structure theory of Petri nets. The method proposed is achieved in two phases. First, from a Petri net model of a given FMS, the authors built an augmented net by adding a "local control place" to each not controlled minimal siphon of the initial net. In spite of the fact that the augmented net is deadlock-free for some classes of FMS, the occurrence of deadlock remains possible. That's why this phase is called "near-prevention" phase. The authors show that the augmented net reaches necessarily an unsafe marking before the occurrence of a deadlock. An unsafe marking is a marking where at least a "local control place" is token-free. So, the second phase of the proposed method consists to falsify as long as possible this necessary condition (unsafe markings) of occurrence of deadlocks in a dynamic way. This phase is implemented by an efficient Petri net controller for which effective conflict transitions are solved according to an appropriate resource allocation policy.