Multicommodity hub-and-spoke network design -

Abstract

Hub-and-spoke is a network architecture with several applications in a number of industries. This type of networks allows the flow of information, people, and products from origin to destination points through central nodes known as hubs. The consolidation of flow at central locations leads to significant reduction in cost due to the economies of scale. Although in practice, several types of commodities flow over such networks, the majority of models in the literature consider single commodity type networks. Dealing with a single commodity is a limitation for such networks, as it is impossible to model multiple types of flow and it does not allow modeling multiple source and destination pairs which is primordial in all sorts of fields. In this research, we present an optimization model for multicommodity hub-and-spoke network design. We present computational testing on a set of instances randomly generated with the Mulgen generator, in addition to a real case network inspired by the French rail network. The results show that small to medium-size networks can be solved within a reasonable computational time. Furthermore, since the major drawback of hub-and-spoke networks is congestion on hubs, we present an extension to the model that includes congestion costs, in order to mitigate the effect of congestion in the network. Since the resulting optimization problem is non-linear, we then present a cutting plane approach based on a piecewise linear approximation as a solution approach.