Constructive heuristics for the canister filling problem

The canister filling problem arises when optimizing the spent nuclear fuel repository in hard rock. It is shown the problem is NP-hard. Constructive heuristics followed by remove and reinsert local optimization are considered. Additionally, some theoretical insight on the algorithm operation is provided. Several variants of the algorithm are compared on random and realistic datasets. The obtained results have shown that constructive heuristics give satisfactory results for both random and realistic input data, although there is still some room for improvements.