In the context of logistics network optimization, the problem of single-assignment multi-hub hub-and-spoke logistics network needs to be solved urgently. This paper constructs a mixed integer linear programming model without capacity restriction and proposes an improved tabu search algorithm. The greedy algorithm optimizes the initial solution, single node exchange constructs the neighborhood, and dynamic tabu table management improves efficiency. Experiments are conducted on 20 city logistics networks. The results showed that a moderate taboo length stabilized the algorithm performance at 0.6–0.8, and a reasonable discount coefficient made the network cost close to 1 after 100 tests, verifying that the algorithm is feasible and efficient. The application of the taboo search algorithm can not only significantly improve the efficiency of logistics resource allocation and reduce related costs but also has obvious advantages over traditional algorithms in terms of solution quality and can find better solutions.