This study focuses on horizontal collaboration and addresses a collaborative multi-center vehicle routing problem (CMCVRP) deriving from the unmanned micro warehouse pattern in new retail in China. In the CMCVRP, suppliers form a coalition to reduce their operating costs and improve service levels. Traditional approaches cannot coordinate the interests of all decision subjects in this process. To fill this gap, an integrated modeling method is proposed considering the objectives at both coalition and partner levels. In the model, two constraints are presented to guarantee the interests of all decision subjects. The constraint at the coalition level defines the acceptable region of solutions to ensure the efficiency of the coalition. And the constraint at the partner level limits the difference among partners’ benefits to avoid excessive profit imbalance in the alliance. After that, a metaheuristic algorithm called the non-dominated sorting genetic algorithm-large neighborhood search is proposed to solve the model. Numerical experiments are conducted on thirty instances with three kinds of logistics networks, and the integrated method is compared with an existing approach (centralized method). The results show that the integrated method has a better performance to coordinate the benefits of all decision-makers when forming the coalition than the centralized method. Also, the impact of the network structures in the collaboration on the performance of the integrated method is investigated, and the conclusions can provide guidance for suppliers to choose their cooperators.