Abstract: Aiming at the problems such as local optimum, low search efficiency and poor real-time obstacle avoidance ability in the path search of traditional AGV path planning algorithms in complex warehouse environments, an AGV dynamic path collaborative planning algorithm integrating improved A^* and DWA is proposed. Firstly, the heuristic function of the traditional A^* algorithm is improved and reconstructed. A dynamic weight coefficient is introduced to enhance the search efficiency, and the global path is streamlined by reducing redundant steering nodes. Secondly, the improved global path key nodes are taken as the current target points of DWA, and the local path evaluation function system is improved. Under the premise of maintaining global optimality, the dynamic obstacle avoidance ability is enhanced to prevent falling into local optimum. Meanwhile, the key point selection strategy was improved, and dynamic windows were used to quickly detect the next target point, significantly enhancing the obstacle avoidance performance and path search efficiency in dynamic environments. The simulation results show that in a complex warehousing environment, compared with the traditional algorithm, the improved fusion algorithm reduces the average path length by 35%, the average search time by 46%, and the number of path turns by 58%. It effectively improves the efficiency of dynamic path planning and real-time obstacle avoidance ability, and has certain superiority and effectiveness.