GVNS for a real-world Rich Vehicle Routing Problem with Time Windows

Rich Vehicle Routing Problems are vehicle routing problems (VRPs) that deal with additional constraints, which aim to better take into account the particularities of real-world applications. They combine multiple attributes, which constitute a complement to the traditional models. This work proposes an adaptive solution method based on metaheuristics for solving a Rich Vehicle Routing Problem with Time Windows. This software has been embedded into the fleet management system of a company in the Canary Islands. The attributes considered by the company are a fixed heterogeneous fleet of vehicles, soft and multiple time windows, customer priorities and vehicle-customer constraints. Furthermore, the company requires the consideration of several objective functions that include travelled distance and time/distance balance. Exact algorithms are not applicable when solving real-life large VRP instances. This work presents a General Variable Neighbourhood Search metaheuristic, which obtains high quality solutions. The computational experiments are presented in four sections, which comprise the parameter setting, the analysis of the effect of the considered attributes, the comparative with the literature for the standard VRP with Time Windows, and the study of the solutions provided by the algorithm when compared with the solutions implemented by the company.