A clustering-routing approach for assigning customers to depots in last mile delivery

Abstract

The distribution of goods in crowded city centres is a major challenge today. To address this problem, two-tier or multi-tier distribution systems have been proposed to consolidate parcels flow in intermediary depots such as for instance urban consolidation centres (UCCs). In this paper, we focus on both assigning customers to depots and evaluating the impact of depots’ location. To address these problems, we propose a two-step approach: first, a k-means clustering algorithm is used to group customers’ demand, secondly, a vehicle routing heuristic evaluates the clusters in terms of number of vehicles (i.e. number of routes) and total distance travelled to serve all customers. Experiments are carried out using this approach on real instances of customers’ demand in the city of Bordeaux. Lastly, we analyze the results of the parcels’ distribution in this city centre and compare two situations in terms of vehicle routing: on the one hand, the situation based on existing and operational depots (i.e. real depots) and, on the other hand, the situation exploiting the depots located in the theoretical centre of the clusters (i.e. k-means depots).