MIP-based heuristics for capacitated lotsizing problems

The complexity of capacitated lotsizing problems is well known. The introduction of setup times and minimum lotsizes makes the problem even more complicated. To handle this complexity, we propose two heuristics based on an iterative procedure. An important feature of our approach is an MIP-based algorithm within a rolling horizon framework. The horizon is partitioned into several sub-horizons over which either a freezing strategy or a simplification strategy can be applied. This allows the relevant MIP model to be reduced, thereby limiting the computational load. To assess the performance of the proposed heuristics, various experiments are reported.