Solving a real world steel stacking problem

We present a real world steel slab stacking problem. The problem features continuous production and retrieval and non-instantaneous crane movements. There are stacking constraints, based on dimensions, weight and temperature of the slabs and temporal constraints based on the casting schedule, the delivery due times as well as the availability of rolling pallets for transport. We present a prioritisation heuristic for the possible crane movements. Based on this heuristic we build a branch and bound solver as well as a greedy look-ahead heuristic. We evaluate the heuristics using randomly generated problem instances of various sizes with the same characteristics as the real world problem and find that the greedy look-ahead heuristic outperforms the branch and bound approach when using realistic time limits.