The value of integrating loading and routing

• Published in: European journal of operational research 2017-02, Vol.257 (1), p.89-105
• Main Authors: Côté, J.F., Guastaroba, G., Speranza, M.G.
• Format: Article
• Language: English
• Subjects: Benchmarks; Integrated approach; Integration in logistics; Loading constraints; Logistics; Orthogonal packing; Routing; Studies; Vehicle routing problems; Worst-case analysis
• ISSN: 0377-2217; 1872-6860
• DOI: 10.1016/j.ejor.2016.06.072

•We consider the problem of integrating routing and loading.•We compare one integrated and three not integrated solution approaches for the 2L-CVRP.•We quantify the potential benefit that can be achieved solving the 2L-CVRP directly.•We analyze the worst-case performance of the not integrated solution approaches.•We show empirically the importance of the integration for this problem. Location-routing, inventory-routing, multi-echelon routing, routing problems with loading constraints are classes of problems that are receiving increasing attention in the scientific community. Problems in these classes generalize classical vehicle routing problems enlarging the decision space to optimize a broader system. The resulting problems are computationally harder to solve but offer opportunities to achieve remarkable additional savings. In this paper, we address the issue of quantifying the potential benefit deriving from tackling directly such complex problems instead of solving the individual problems in a not integrated manner. To this aim, we consider as a proof of concept the Capacitated Vehicle Routing Problem (CVRP) with Two-dimensional Loading constraints (2L-CVRP), a variant of the CVRP where rectangular-shaped items have to be delivered to customers and loading constraints have to be satisfied. We consider the 2L-CVRP in an integrated manner and compare the solutions with those obtained from three not integrated approaches based on addressing separately the routing and the loading problems. The importance of an integrated approach for the 2L-CVRP is validated through the study of the worst-case performance of the not integrated approaches, and conducting computational experiments on benchmark and new instances.