In the literature, the hub location problem is commonly modeled as hub median or hub center problems (Meyer, Ernst, & Krishnamoorthy, 2009). The main objective of hub median problems is to minimize total transportation cost. If there is a maximum limit on the number of hubs, it is called the p-hub median problem. In hub center problems on the other hand, the objective is to minimize maximum distance (cost) between Origin–Destination (O/D) pairs. Next to these two formulations, hub location problems are sometimes modeled as hub covering problems where the objective function is to maximize the total number of served spoke nodes. For an extensive study on hub location problems, we refer to Alumur and Kara (2008).
Fig. 1 shows that there are relatively few papers studying hub covering and hub center problems, compared to hub median problems. Where maximizing market share and customer recognition is the prime goal, hub covering problems offer the best approach to model them. In comparison, hub center problems are suitable for designs where immense worst-case O/D distance is not desirable, especially in time-sensitive delivery systems. These problems are interesting to study.
One of the assumptions generally made in hub location problems is that the interhub network is a complete graph, but the spoke nodes are not always interconnected. Direct shipment between spoke pairs is also not allowed and the flow of cargo traverses at most two hubs. In practice, especially on an international scale, such a rigid structure is unlikely. Direct shipment by trucks is commonly used, and complete multimodal networks are not feasible. These facts bring many modeling and methodological challenges which need more investigation. A considerable part of the literature has a direct shipment option in their model. However, only Rodríguez-Martín and Salazar-González, 2008 and Alumur et al., 2012a assume an incomplete underlying network.