Pablo Azurduy, Marcel Goic, Cristián E. Cortés, Nicolás Aramayo
Abstract
In recent years, the development of information technologies has led to significant changes in the design of the logistics systems that companies use to distribute products to their customers. Among the most relevant changes is the emergence of digital platforms that connect firms that need to deliver their products with drivers who can handle last-mile delivery. One of the central attributes of these platforms is the lack of a contractual relationship with potential drivers. Therefore, when a platform offers a possible route, drivers will accept only those routes they find more attractive. This logic drives a reevaluation of traditional solution approaches to address the resulting vehicle routing problem (VRP), aiming to generate solutions that reduce transportation costs and offer more attractive routes for drivers. In this paper, we propose a conceptual framework to address the VRP with noncontractual drivers (VRPND), consisting of an integrated solution scheme that learns from drivers’ preferences and incorporates them to generate routes likely to be accepted. By applying our approach to a real-world problem of a digital logistics platform, we find that we can reduce acceptance times, demonstrating a better alignment between the proposed routes and driver preferences.
