Journal of Transport Geography 1994 2(1)31-40 The hub network design problem A review and synthesis Morton E.O'Kelly Department of Geography,The Ohio State University,Columbus,Ohio 43210,USA Harvey J.Miller Department of Geography,University of Utah,Salt Lake City,Utah 84112.USA Hubs,or central trans-shipment facilities,allow the construction of a network where large numbers of direct connections can be replaced with fewer,indirect connections.Hub-and- spoke configurations reduce and simplify network construction costs,centralize commodity handling and sorting,and allow carriers to take advantage of scale economies through consolidation of flows.Such networks have widespread application in transportation.This paper presents a structured review of research on the hub network design problem.Three critical design questions need to be considered:(a)are the nodes in the network assigned exclusively to a single hub?(b)are direct node-to-node linkages permitted to bypass the hub facilities?and,(c)are the hub facilities fully interconnected?The nature and difficulty of the hub network design problem depends on the analyst's judgement with respect to these questions.We review analytical research papers,and give brief empirical examples of eight different network design protocols. Keywords:hub and spoke,network design.location Flows of people,commodities,information and approaches to the problem,it is difficult to form any energy all require a complex network of interlinkages generalizations.Indeed,there exists a disconcerting between origins and destinations.A special kind of number of definitions and ideas about what con- network,namely,the hub network is designed for stitutes a hub.The following paragraphs discuss servicing human,commodity or information flows briefly the various concepts of hub which have been between multiple origins and destinations,ie,the used in the operations research literature,in air many-to-many distribution problem.Hubs,or central passenger transportation and in air package trans-shipment facilities,allow the construction of a delivery network where direct connections between all origin In the case of the early literature on management and destination pairs can be replaced with fewer, science and operations research,the concept of hub indirect connections.These configurations reduce seems to have been synonymous with a central and simplify network construction costs,centralize warehouse or facility.(See Minas and Mitten [1958] commodity handling and sorting,and allow carriers where a model for scheduling truck movements in to take advantage of scale economies through and out of a depot or hub is presented without any consolidation of flows (Chestler,1985;Devany and notion of sorting or throughput.)Thus,a hub is in Garges,1972;Kanafani and Ghobrial,1985;Toh essence a warehouse or a central depot,which is and Higgins,1985). located at the centre of a set of demand regions Hub-and-spoke networks are applicable to many Conversely,Goldman (1969)analysed what is different types of transport problem.Examples of actually a hub facility,but referred to it as a 'center'. hub-and-spoke systems include:airline passenger As noted by Campbell (1991a)Goldman located carriers (Shaw,1993);overnight package delivery centres to minimize the sum of transport costs over a services (Chestler,1985);and rail sorting yards set of origin-destination pairs,and so formulated a (Bodin et al,1980).While these diverse applications general multiple-hub assignment problem. are very well known,the prospects for a compre- In air passenger transportation,as defined by the hensive model for hub network optimization are US Federal Aviation Administration (FAA),the limited at the moment.There are so many different term 'hub'is not based on the hub-and-spoke 0966-6923/94/010031-10 1994 Butterworth-Heinemann LtdJournal of Transport Geography 19942(1) 31-40 The hub network design problem A review and synthesis Morton E. O'Kelly Department of Geography, The Ohio State University, Columbus, Ohio 43210, USA Harvey J. Miller Department of Geography, University of Utah, Salt Lake City, Utah 84112, USA Hubs, or central trans-shipment facilities, allow the construction of a network where large numbers of direct connections can be replaced with fewer, indirect connections. Hub-and￾spoke configurations reduce and simplify network construction costs, centralize commodity handling and sorting, and allow carriers to take advantage of scale economies through consolidation of flows. Such networks have widespread application in transportation. This paper presents a structured review of research on the hub network design problem. Three critical design questions need to be considered: (a) are the nodes in the network assigned exclusively to a single hub? (b) are direct node-to-node linkages permitted to bypass the hub facilities? and, (c) are the hub facilities fully interconnected? The nature and difficulty of the hub network design problem depends on the analyst's judgement with respect to these questions. We review analytical research papers, and give brief empirical examples of eight different network design protocols. Keywords: hub and spoke, network design, location Flows of people, commodities, information and energy all require a complex network of interlinkages between origins and destinations. A special kind of network, namely, the hub network is designed for servicing human, commodity or information flows between multiple origins and destinations, ie, the many-fo-many distribution problem. Hubs, or central trans-shipment facilities, allow the construction of a network where direct connections between all origin and destination pairs can be replaced with fewer, indirect connections. These configurations reduce and simplify network construction costs, centralize commodity handling and sorting, and allow carriers to take advantage of scale economies through consolidation of flows (Chestler, 1985; Devany and Garges, 1972; Kanafani and Ghobrial, 1985; Toh and Higgins, 1985). Hub-and-spoke networks are applicable to many different types of transport problem. Examples of hub-and-spoke systems include: airline passenger carriers (Shaw, 1993); overnight package delivery services (Chestler, 1985); and rail sorting yards (Bodin et ai, 1980). While these diverse applications are very well known, the prospects for a compre￾hensive model for hub network optimization are limited at the moment. There are so many different 0966-6923/94/010031-10 © 1994 Butterworth-Heinemann Ltd approaches to the problem, it is difficult to form any generalizations. Indeed, there exists a disconcerting number of definitions and ideas about what con￾stitutes a hub. The following paragraphs discuss briefly the various concepts of hub which have been used in the operations research literature, in air passenger transportation and in air package delivery. In the case of the early literature on management science and operations research, the concept of hub seems to have been synonymous with a central warehouse or facility. (See Minas and Mitten [1958] where a model for scheduling truck movements in and out of a depot or hub is presented without any notion of sorting or throughput.) Thus, a hub is in essence a warehouse or a central depot, which is located at the centre of a set of demand regions. Conversely, Goldman (1969) analysed what is actually a hub facility, but referred to it as a 'center'. As noted by Campbell (1991a) Goldman located centres to minimize the sum of transport costs over a set of origin-destination pairs, and so formulated a general multiple-hub assignment problem. In air passenger transportation, as defined by the US Federal Aviation Administration (FAA), the term 'hub' is not based on the hub-and-spoke