A Monte-Carlo-rollout algorithm for stock control

The application of optimization in industrial processes is faced with many challenges. One of the main challenge is the possible inaccuracy of information. In contrast to mathematical optimization theory, information is not completely known a priori. Often information can only be estimated or changes over time. Another challenge is the need of a decision in real time. Both points are relevant for a control of a flexibly designed in-plant block storage. The schedule plan for storages and removals should be able to adapt quickly to changes. In this paper an algorithmic approach is presented which is able to react on dynamic and uncertain changes due to the production process. To this end, optimization algorithms are implemented within a rolling planning process, so it is possible to respond to updated information by adapting the current plan. A novel optimization method is developed to generate cost effective and robust solutions by looking ahead into the future.