A large neighborhood search-based matheuristic for a fuel distribution problem with time windows

In this work, we propose a hybrid approach based on large neighborhood search (LNS) and logic-based Benders decomposition (LBBD) to solve a fuel distribution problem with heterogeneous fleet and compartmentalized vehicles in a more scalable way. The approach combines a randomized gluttonous constructive heuristic with a multistart strategy to generate viable and diversified initial solutions, and an iterative cycle of destruction and reconstruction in which reconstruction is performed by a restricted version of LBBD, defined on reduced vehicle sets and trips. This strategy allows you to reconstruct partial solutions by accurately solving smaller sub-problems, significantly reducing memory cost. The computational experiments show that the proposed method consistently achieves the best known solutions, or values very close to them, and expands the size of the instances that can be treated in practice, maintaining good solution quality.