An ALNS metaheuristic for a dynamic hierarchical facility location problem with modular capacities
dynamic facility location, modular capacities, hierarchical facilites, metaheuristic, large neighborhood search.
The necessity to design real supply chains using mathematical models in the facility location problems context led to the rise of very elaborate models. Literature about these problems in operations research is vast, but there is still a lot to be explored. Since the appearance of the area in the mid-twentieth century, new features are being incorporated to classic models arising several new classes of problems. This study suggests the unification of some of these features that have never been explored together in a new and unique model of a facility location-allocation problem. The resulting model is capable of handle with up to two levels of hierarchical facilities, multiple commodities, multiple allocations between facilities and customers, and modular adjustments on the capacity of the facilities, all in a planning horizon. Since facility location problems are difficult to be solved by exact methods, in the second part of this work we developed an Adaptive Large Neighborhood Search (ALNS) metaheuristic, which should be able to solve the suggested problem and other sub-problems derived from it. The ALNS framework is relatively new in the literature, and it has been gaining attention for its flexibility when dealing with problems in which the set of solutions in the search space is too big to be exploited by simple heuristics alone. Its adaptive structure allows a collection of heuristics to be implemented but only those that are performing better are called with more frequency. As these are independent heuristics, different neighborhoods can be exploited at each iteration of the algorithm. Preliminary tests with benchmarking instances of a sub-problem found in literature suggest that the metaheuristic implemented has the potential to be competitive in terms of both time resolution and quality solution when compared to specialized heuristics to this sub-problem