Stochastic production scheduling to meet demand forecasts

Abstract

Production scheduling, the problem of sequentially configuring a factory to meet forecasted demands, is a critical problem throughout the manufacturing industry. We describe a Markov decision process (MDP) formulation of production scheduling which captures stochasticity, while retaining the ability to construct a schedule to meet demand forecasts. The solution to this MDP is a value function, specific to the current demand forecasts, which can be used to generate optimal scheduling decisions online. We then describe an industrial application and a reinforcement learning method for generating an approximate value function in this domain. Our results demonstrate that in both deterministic and noisy scenarios, value function approximation is an effective technique.