Production planning and control with discrete lotsizing and a rolling horizon

We consider a production system with discrete lotsizing as it is given in the automotive industry. At any point of time only orders in the near future are known, but still the production has to be planned - leading to rolling horizons. Decisions that are made today should avoid bottlenecks in stock and capacity that will handicap future production. First we discuss approaches used in computing. Second we propose an approach that perform with good competivity compared to an optimal algorithm that knows the entire sequence of orders in advance.