Combining autonomous delivery robots and traditional vehicles with public transportation infrastructure in last-mile distribution

The recent boom of e-commerce and the resulting concern about emissions increase are promoting the development of alternative paradigms for last-mile delivery, involving greener vehicles, such as air drones and autonomous delivery robots, which are also less sensitive to traffic congestion. In this paper, we study a last-mile distribution problem in which a heterogeneous fleet composed of both traditional vehicles and autonomous delivery robots is used to service a large number of customer requests in an urban area. The delivery robots use stations of the public transportation network as their depots. The problem amounts to building both van and delivery-robot routes in such a way as to minimize the distribution cost, meeting the constraints imposed by the different ranges and capacities of both vehicles. At the same time, another decision involves choosing at which stations of the public transportation network the delivery robots are based. For this purpose, we develop two variants of a matheuristic scheme based on the solution of a hybrid set-partitioning and location-based model. We perform thorough computational experiments on instances inspired by the problem of distributing drugs to pharmacies in the urban area of Rome (Italy) and show that, when the instance size grows, the proposed solution approach allows us to obtain significant cost reductions compared to both a straightforward solution of a mathematical model and a classical metaheuristic scheme. In addition, we show that combining autonomous delivery robots with traditional vehicles allows us to achieve important cost and emission reductions with respect to a traditional delivery scheme.