Computation of Optimal Flow Control Policies of a Manufacturing System with Multiple Production Rates

We address the problem of nding optimal hedging points of a production ow controller, in the sense of Bielecki and Kumar (1988). We model the dynamics of the controller by using generalized semi-Markov processes (GSMP); the GSMP framework lets us work with multiple states, and not only with exponential distribution but with fairly general distributions. Utilizing the GSMP representation, the sensitivity estimates of cost functions, consisting of inventory and backlog costs, with respect to hedging points are computed. We then discuss how these sensitivity estimates are used together with sample-path optimization , a recent simulation-based optimization method, for nding optimal hedging points.