针对未来无人机（UAV）规模化应用和密集化飞行带来的安全问题，面向多元化空域、高密度流量和高动态演化交织形成的复杂低空系统，研究了复杂低空无人机飞行冲突网络建模与精细管理问题。首先，采用Gilbert-Johnson-Keerthi（GJK）算法，构建了基于冲突连边动态识别的无人机飞行冲突网络模型，以平均冲突数量、平均运行风险和平均聚集系数等指标表征了无人机飞行冲突网络特征参数；然后，基于空间栅格编码理论，提出了低空空域系统栅格剖分和编码设计方法，建立了离散空域环境基于数字栅格的无人机飞行冲突探测算法，实现了无人机飞行冲突的精细探测；最后，以无人机冲突数量最少和运行成本最低为优化目标，构建了考虑优先级的无人机冲突解脱多目标混合整数优化模型，设计了基于非支配排序遗传算法（NSGA-Ⅱ）算法的无人机冲突解脱智能算法，实现了多任务场景和不同优先级下无人机飞行冲突的最优解脱。仿真实验表明，与传统的基于三维坐标的冲突探测方法相比，提出的基于数字栅格的冲突探测方法可将探测总时间降低78.4%，并可将多机型混杂场景下的无人机冲突探测效率由“指数增长级”降至“线性增长级”；以平均冲突风险等级作为解脱优先级原则时，无人机冲突解脱模型可得到性能更优的非支配解，与不考虑优先级相比，可将平均每架无人机高度层、水平航迹和速度的调整值分别减少32.8%、21.4%和14.6%。经验证，本文提出的无人机飞行冲突精细管理方法是有效的，可为低空空域安全管理与无人机风险防控提供理论支撑和方法指导。

The large-scale application and intensive flight of Unmanned Aerial Vehicle （UAV） in the future bring about safety problems. In view of the complex low-altitude system formed by the interweaving of diversified airspace， high-density traffic and high dynamic evolution， the problem of network modeling and refined management of complex low-altitude UAV flight conflicts is studied. Firstly， the Gilbert-Johnson-Keerthi （GJK） algorithm is used to construct a UAV flight conflict network model based on the dynamic identification of conflicting edges， and the characteristic parameters of the UAV flight conflict network are characterized by the average number of conflicts， the average operation risk and the average clustering coefficient. Then， based on the spatial raster coding theory， a raster segmentation and coding design method for low-altitude airspace system is proposed， and a digital raster-based UAV flight conflict detection algorithm for discrete airspace environment is established. Finally， a multi-objective mixed integer optimization model for UAV conflict resolution considering priority is constructed with the optimization objectives of the minimum number of UAV conflicts and the lowest operation cost. Moreover， an intelligent UAV conflict resolution algorithm based on Non-Dominated Sorting Genetic Algorithm-Ⅱ（NSGA-Ⅱ） algorithm is designed to realize the optimal resolution of UAV flight conflicts under multi-task scenarios and different priorities. The optimal resolution of UAV flight conflicts under multi-mission scenarios and different priorities is then achieved. Simulation experiments show that compared with the traditional 3D coordinate-based conflict detection method， the proposed digital raster-based conflict detection method can reduce the total detection time by 78.4%， and reduce the UAV conflict detection efficiency from “exponential growth level” to “linear growth level” in multi-model hybrid scenarios. When the average conflict risk level is used as the deconfliction priority principle， the UAV conflict deconfliction model can obtain non-dominated solutions with better performance and reduce the average adjustment values of altitude layer， horizontal trajectory and speed per UAV by 32.8%， 21.4% and 14.6% compared with the principle without considering the priority. It is verified that the fine management method of UAV flight conflict proposed in this paper is effective and can provide theoretical support and methodological guidance for the safety management of low-altitude airspace and UAV risk prevention and control.