摘要： 为解决复杂堆场中多无人机巡检面临的目标密集分布、遮挡关系复杂、标签识读可靠性不足及多无人机路径冲突等问题，提出一种多无人机巡检任务分配与路径规划方法。综合考虑链路传播、天线方向、极化失配、阴影衰落及堆叠遮挡等因素，建立兼顾RFID可靠识读与多无人机安全飞行约束的巡检模型，将问题分解为覆盖约束下的可靠悬停位置优化，以及多无人机任务分配与无冲突路径规划两个子问题。设计基于势博弈的可靠悬停位置优化算法，以截断型势函数度量未可靠覆盖标签的概率增益，通过多初始最佳响应动态搜索生成满足覆盖率约束且规模紧凑的悬停位置集合。基于该集合，设计基于冲突反馈与安全区间规划的任务分配与路径规划算法（CF-SIPP），以系统完工时间、总时间代价、负载均衡度和冲突惩罚构造综合目标，通过任务转移、任务交换和访问序列调整优化任务方案，并将安全区间路径规划产生的冲突信息反馈至任务调整过程，实现任务优化与无冲突路径生成。仿真结果表明，在不同无人机数量和标签规模实验中，CF-SIPP相较ECBS-TS的系统完工时间分别降低6.33%和3.39%，总时间代价分别降低2.69%和2.01%；在复杂工况和路径阻塞条件下，系统完工时间分别降低6.44%和6.71%，总时间代价分别降低6.75%和6.93%。消融实验表明，任务转移是降低系统完工时间的主要局部搜索策略，冲突反馈在高冲突场景下作用更为明显；参数敏感性分析表明，遮挡附加损耗是影响巡检性能的关键因素。显著性检验结果表明，CF-SIPP在不同任务规模和环境约束下均具有稳定的性能优势。所提方法能够在满足标签可靠覆盖和多无人机安全间隔约束的同时，有效降低系统完工时间和总时间代价，并具有较好的复杂工况适应性、规模扩展性和运行稳定性。

Abstract: To address the challenges of dense target distribution, complex occlusion, insufficient tag-reading reliability, and path conflicts among multiple UAVs in complex yards, this paper proposes a task allocation and path planning method for multi-UAV inspection. Considering link propagation, antenna directivity, polarization mismatch, shadow fading, and additional loss caused by stacked occlusions, an inspection model incorporating reliable RFID reading and multi-UAV safe-flight constraints is established, and the problem is decomposed into two subproblems: reliable hovering-position optimization under coverage constraints, and multi-UAV task allocation with conflict-free path planning. A potential-game-based reliable hovering-position optimization algorithm is designed, in which a truncated potential function is used to quantify the probability gain of tags that have not yet been reliably covered, and multi-start best-response dynamics are employed to generate a compact hovering-position set satisfying the coverage constraint. Based on this set, a task allocation and path planning algorithm based on conflict feedback and safe interval planning (CF-SIPP) is developed. A composite objective incorporating makespan, total time cost, load balancing, and conflict penalties is constructed. The task plan is refined through task transfer, task exchange, and visit-sequence adjustment, while conflict information generated by safe interval path planning is fed back to the task adjustment process, thereby enabling task optimization and conflict-free path generation. Simulation results show that, compared with ECBS-TS, CF-SIPP reduces the makespan by 6.33% and 3.39% in experiments with different numbers of UAVs and different tag scales, respectively, while reducing the total time cost by 2.69% and 2.01%, respectively. Under complex operating conditions and path-blocking conditions, the makespan is reduced by 6.44% and 6.71%, respectively, and the total time cost is reduced by 6.75% and 6.93%, respectively. Ablation experiments indicate that task transfer is the principal local-search strategy for reducing the makespan, while conflict feedback plays a more significant role in high-conflict scenarios. Parameter sensitivity analysis shows that additional occlusion loss is a key factor affecting inspection performance. Significance tests further demonstrate that CF-SIPP provides stable performance advantages under different task scales and environmental constraints. The proposed method can effectively reduce the makespan and total time cost while satisfying reliable tag coverage and multi-UAV safety-separation constraints, and exhibits good adaptability to complex operating conditions, scalability, and operational stability.