Low-Cost Deadlock-Avoidance Policies For Flexible Production Systems

Today, the demand for a larger variety of goods requires more flexible production systems, able simultaneously to carry out production of changing characteristics and to process pieces of different types. Flexible production systems exhibit a high degree of resource sharing, which can lead to congestion situations, like deadlocks. Deadlocks are highly undesirable events, as a set of parts remains in "circular wait" condition indefinitely. To avoid deadlock, suitable control policies governing resource acquisition and release are necessary. This paper analyzes three feedback policies for deadlock avoidance, comparing both their costs for offline and online computations and the restrictions they impose on system performance. One of these policies is the least restrictive control rule that one can find, but it is applicable only to production systems satisfying some constraints. In this context, the paper determines a class of production systems for which such a policy can be applied. The discrete event simulation of some case studies confirms the theoretical results discussed in the paper.