Um algoritmo hybrid genetic search para o problema de roteamento de veículos com armários de encomendas

Abstract

Parcel lockers have become an increasingly relevant alternative for last-mile logistics, motivating the study of the Vehicle Routing Problem with Parcel Lockers (VRPPL), a recent variant of the classic Vehicle Routing Problem (VRP) that integrates routing decisions with customer-to-locker allocation. This work evaluates the application of a Hybrid Genetic Search (HGS)-based approach to the VRPPL, implemented using the PyVRP framework and adapted through a modeling strategy based on mutually exclusive customer groups and a local search operator designed to consolidate visits to parcel lockers. The proposed approach was evaluated on benchmark instances from the literature and compared with methods based on Simulated Annealing and Iterated Greedy. Experiments demonstrate that the method achieves competitive performance, showing clear advantages over Simulated Annealing regarding runtime, and solution quality comparable to that obtained by the Iterated Greedy, indicating that the HGS represents a viable and robust alternative for solving the VRPPL.