Otimização de rotas de transporte com múltiplos drones: Proposta utilizando machine learning, meta-heurísticas e solução exata

Abstract

The increasing utilization of drones for last-mile deliveries, motivated by the reduction of delivery time and maintenance costs, as well as by the mitigation of CO2 emissions, demands robust solutions for the multiple drone transport route optimization problem (Covering Salesman Problem with Multiple Drones – CSPMD). This article proposes a multilayer solution to this problem, in an approach that integrates Machine Learning (ML), metaheuristic, and dynamic programming. The obtained results indicate that the use of specific metaheuristics, in the context of this multilayer framework, constitutes a promising strategy for solving the CSPMD. The performance evaluation of these metaheuristics indicated that the Genetic Algorithm outperformed in small instances, achieving a cost up to 0.49% lower, while Simulated Annealing proved superior in large instances, with a cost up to 1.93% lower.