电子商务仓储环境下基于改进A*算法的AGV路径规划研究

Research on AGV Path Planning Based on Improved A* Algorithm in E-Commerce Warehousing Environment

摘要: 针对电子商务仓储环境下AGV路径规划的冲突问题，提出一种基于坐标保留表和冲突分类的无冲突多目标算法CF-MOWVRP (Conflict-Free Multi-Objective Warehousing Vehicle Routing Problem)。该算法首先构建鱼骨式栅格地图，模拟货架、拣选台和动态任务环境；其次改进A*算法，支持4方向搜索和路径缓存，然后建立多目标优化模型，最小化总运输距离、最大单次距离和冲突等待时间，并通过遗传操作、偏好随机策略和强化学习求解Pareto前沿；最后利用折线优化和二次Bezier曲线平滑路径，提高AGV运动平滑度。仿真结果表明：在30个动态任务场景下，相比传统A*，节点减少33%、转弯减少50%、完成时间降低11%。AGV增至10辆、任务增至60个、障碍比例增至0.45时，完成时间均降低约17%，验证了算法在高并行、高密度和复杂环境中的鲁棒性。

Abstract: Addressing the conflict issue of AGV path planning in the e-commerce warehousing environment, a conflict-free multi-objective algorithm CF-MOWVRP (Conflict-Free Multi-Objective Warehousing Vehicle Routing Problem) based on coordinate retention table and conflict classification is proposed. This algorithm first builds a fishbone-like grid map to simulate shelves, picking stations and dynamic task environments; then improves the A* algorithm to support four-direction search and path caching; then establishes a multi-objective optimization model to minimize total transportation distance, maximum single distance and conflict waiting time, and solves the Pareto frontier through genetic operations, preference random strategies and reinforcement learning; finally, it optimizes the path with polyline optimization and quadratic Bezier curve smoothing to improve the smoothness of AGV movement. The simulation results show that in 30 dynamic task scenarios, compared to traditional A*, nodes are reduced by 33%, turns are reduced by 50%, and completion time is reduced by 11%. When the number of AGVs increased to 10, the number of tasks increased to 60, and the obstacle ratio increased to 0.45, the completion time decreased by about 17%, verifying the robustness of the algorithm in high parallel, high-density, and complex environments.