Neuroevolutionary Inventory Control in Multi-Echelon Systems
Stochastic inventory control in multi-echelon systems poses hard problems in optimisation under uncertainty. Stochastic programming can solve small instances optimally, and approximately solve large instances via scenario reduction techniques, but it cann
- PDF / 227,923 Bytes
- 12 Pages / 430 x 660 pts Page_size
- 6 Downloads / 189 Views
2
Cork Constraint Computation Centre, Ireland Operations Management Division, Nottingham University Business School, Nottingham, UK 3 Logistics, Decision and Information Sciences Group, Wageningen UR, The Netherlands 4 Faculty of Computer Science, Izmir University of Economics, Turkey [email protected], [email protected], [email protected], [email protected]
Abstract. Stochastic inventory control in multi-echelon systems poses hard problems in optimisation under uncertainty. Stochastic programming can solve small instances optimally, and approximately solve large instances via scenario reduction techniques, but it cannot handle arbitrary nonlinear constraints or other non-standard features. Simulation optimisation is an alternative approach that has recently been applied to such problems, using policies that require only a few decision variables to be determined. However, to find optimal or near-optimal solutions we must consider exponentially large scenario trees with a corresponding number of decision variables. We propose a neuroevolutionary approach: using an artificial neural network to approximate the scenario tree, and training the network by a simulation-based evolutionary algorithm. We show experimentally that this method can quickly find good plans.
1 Introduction In the area of optimisation under uncertainty, one of the most mature fields is inventory control. This field has achieved excellent theoretical and practical results using techniques such as dynamic programming, but some problems are too large or complex to be solved by classical methods. Particularly hard are those involving multi-echelon systems, in which multiple stocking points form a supply chain. In such cases we may resort to simulation-based methods. Simulation alone can only evaluate a plan, but when combined with an optimisation algorithm it can be used to find near-optimal solutions (or plans). This approach is called simulation optimisation (SO) and has a growing literature in many fields including production scheduling, network design, financial planning, hospital administration, manufacturing design, waste management and distribution. It is a practical approach to optimisation under uncertainty that can handle problems containing features that make them difficult to model and solve by other methods: for example non-linear constraints and objective function, and demands that are correlated or have unusual probability distributions.
B. Hnich is supported by the Scientific and Technological Research Council of Turkey (TUBITAK) under Grant No. SOBAG-108K027. This material is based in part upon works supported by the Science Foundation Ireland under Grant No. 05/IN/I886.
F. Rossi and A. Tsoukis (Eds.): ADT 2009, LNAI 5783, pp. 402–413, 2009. c Springer-Verlag Berlin Heidelberg 2009
Neuroevolutionary Inventory Control in Multi-Echelon Systems
403
SO approaches to inventory control are typically based on policies known to be optimal in certain situations, involving a small number of reorder points and reorder
Data Loading...
