Deadlock is a critical problem in the control of Automated Flexible Manufacturing Systems. Namely, when a deadlock occurs, the flow of jobs is permanently inhibited so that operations on parts may not be performed. That is why an efficient control policy must avoid deadlock without imposing unnecessary restrictions on part loading and routing. In this paper we generalize the results of a previous companion work dealing with deadlock avoidance in systems containing single capacity resources. Using the same theoretic framework, we rigorously characterize the deadlock occurrence in multiple resource systems. Digraph theory is still effective for deriving necessary and sufficient conditions characterizing highly undesirable situations (second level deadlocks) that inevitably evolve to circular waits in the near future. The results of the analysis let us introduce some control laws (named restriction policies) that avoid deadlocks in multiple-resource systems by inhibiting or enabling some properly identified events. Finally, the paper discusses and compares the computation costs of the proposed restriction policies.
Deadlock Avoidance, discrete event dynamical systems, FMS control