In modern warfare, the demand for ammunition and containerized cargo supply guarantee has the characteristics of suddenness, large quantities and time sensitive. To solve the problem of task allocation for rapid outbound of warehouse under high-intensity combat conditions, a mathematical model for collaborative task allocation of multiple AGVs was established. An improved discrete pigeon-inspired optimization(IDPIO) was proposed, which introduced a probability-based mutation process in the individual iteration process of the compass operator. The global search performance of the improved discrete pigeon-inspired optimization(IDPIO) was enhanced by adjusting the cross-learning rate of the compass operator, which can balance between convergence speed and global search performance. Then, an optimized collaborative task allocation solution for multiple AGVs is obtained. Finally, a simulation with a demand for 20 goods to be moved out by 4 AGVs is carried out. The result shows that compared with the discrete pigeon-inspired optimization (DPIO), the new method has advantages in convergence speed and global search performance.