On the complexity of optimal deadlock avoidance in flexible manufacturing systems

The problem of the manufacturing system deadlock has received considerable attention recently, since its effective solution is a prerequisite to the implementation of large-scale flexibly automated discrete-part manufacturing systems (FMS). Its difficulty arises from the fact that in the FMS context, the computation of the optimal deadlock avoidance policy (DAP) is NP-hard. This paper proves that for a large subclass of FMS, with significant practical relevance, the optimal DAP is polynomially tractable in real-time. The implications of this result to broader FMS classes are also considered.