An Approximate Dynamic Programming Approach for a Product Distribution Problem

This paper proposes an approximate dynamic programming-based method for optimizing the distribution operations of a company manufacturing a certain product in numerous plants and distributing it to different regional markets for sale. The production at each plant and in each time period follows a nonstationary stochastic process. The product can be stored at the plants or shipped to a regional market to serve the random demand. The proposed solution method formulates the problem as a dynamic program and uses approximations of the value function. We develop a tractable procedure to compute the marginal worth of a unit of inventory at a production plant. This quantity is used to update the value function approximations in an iterative improvement scheme. We numerically show that our method yields high quality solutions.