Job Shop Production Planning under Uncertainty: A Monte Carlo Rollout Approach

The Monte Carlo Rollout method (MCR) is a novel approach to solve combinatorial optimization problems with uncertainties approximatively. It combines ideas from Rollout algorithms for combinatorial optimization and the Monte Carlo Tree Search in game theory. In this paper the results of an investigation of applying the MCR to a Scheduling Problem are shown. The quality of the MCR method depends on the model parameters, search depth and search width, which are strong linked to process parameters. These dependencies are analyzed by different simulations. The paper also deals with the question whether the Lookahead Pathology occurs.