Since optimizing the scheduling of the storage and retrieval path of large-scale automated warehouses is difficult， based on the storage and retrieval scheduling model of dual-end double stacker cranes， an improved anti-collision principle is proposed to avoid collisions when dual stackers are running at the same time. The optimal anti-collision boundary inspection mechanism is used to divide the best working area for the two stackers. An improved genetic algorithm （NIGA） is proposed to adjust the evolution structure of the genetic algorithm according to the degree of concentration and dispersion of the population fitness value. Thereby it can effectively improve the convergence speed and break the limit of the local optimum. The NIGA algorithm is used to optimize the scheduling path of the dual stackers， and the improved anti-collision principle and the optimal anti-collision boundary inspection mechanism are embedded in each iteration of the algorithm. Finally， the optimal solution of the two-end dual stacker warehousing is obtained. Simulation results show that this strategy can effectively prevent the collision of two stackers， and significantly improve the efficiency of goods in and out of large warehouses.