Due to the increasing number of automated guided vehicles (AGVs) in the multi-AGV system and the limitation of working environment, path conflicts often occur in the working process of AGVs, which affects the working efficiency of the multi-AGV system. Thus, a optimization method by arranging the AGVs' traffic sequence is proposed in this paper. First, an AGV working map is reconstructed with graph theory, and then the corresponding collision avoidance rules are formulated for different types of conflicts. In multi-AGV system, each collision avoidance decision has an impact on the efficiency of the system, so it is crucial to adopt appropriate decisions. To optimize the decisions, the system fitness of different collision avoidance decisions are calculated based on the global state of the system, and the particle swarm optimization (PSO) algorithm is used to optimize the decisions. Furthermore, the PSO algorithm is improved by planning the direction of particle motion in the solution space and introducing mutation operation, so as to improve the search ability of the particle in the solution space. To verify the feasibility and effectiveness of the improved particle swarm optimization (IPSO) algorithm, an experiment system is built based on.NET platform. Results show that the IPSO algorithm than the traditional algorithms experimental performs better. The IPSO algorithm can effectively reduce congestion caused by path conflict and enhance the efficiency of the multi-AGV system.