A Multi-Start Algorithm and a Large Neighborhood Search for a Maritime Inventory Routing Problem.

Maritime Inventory Routing Problem (MIRP) is a challenging combinatorial problem in which decision solutions evolve vessels route and schedule, besides the inventory management at the ports along a limited planning horizon. Common solution approaches for solving MIRPs use matheuristics combining heuristic elements and mathematical programming for obtaining high-quality solutions in relatively short processing time. The performance of such approaches can be compromised if the problem size grows considerably. To provide an alternative approach for solving larger MIRPs, we propose a multi-start metaheuristic capable of producing several solutions for the problem in short processing time. Additionally, we developed a large neighborhood search for improving a subset of the generated solutions, which are then used to build a reduced mixed-integer problem which is solved by a mathematical solver. Computational results demonstrated that our algorithm can obtain solutions in short processing time, compared to the current solutions methods, and it can solve larger problem instances, in which no solutions were known.