Abstract: In order to cope with the highly decentralized and fluctuating distribution of express transportation volume, a hybrid hub-and-spoke transportation network combining a hub-and-spoke network and direct-connected routes is proposed. On the premise of making full use of the established express service nodes, direct routes are established between non-hub cities to effectively reduce transportation cost. A mixed integer linear programming (MILP) model for the hybrid hub-and-spoke express transportation network is built, with a cost function comprised of transportation cost, loading and unloading cost and holding cost as the optimization objective. A modified genetic simulated annealing algorithm is developed to solve the model: genetic algorithm determines the locations of hubs and the allocations of non-hub nodes, while simulated annealing algorithm solves the direct connections between non-hub nodes. The proposed model and algorithm are applied to a test case with 30 nodes based on real express data in China. Compared with the traditional hub-and-spoke mode, the average cost is reduced by 32.2%, and the gap between the heuristic solution and the optimal solution is less than 5%. In addition, sensitivity analysis shows that trucks with a capacity of 1 000 kg are the optimal transportation modes for redesigning transportation plans, which provides theoretical basis and practical reference for express companies that currently use hub-and-spoke networks to improve transportation routes.

Key words: logistics engineering, hybrid hub-and-spoke network, genetic-simulated annealing algorithm, express transportation, distribution centers, direct-connected network