研究了单架无人作战飞机(UCAV)攻击多个地面目标的三维轨迹规划问题。首先,将问题形式化为一类特殊的旅行商问题(TSP),即带动力学约束的邻域访问TSP问题(DCTSPN)。其次,针对规划空间维度过高、搜索代价过大的问题,提出了一种基于概率路标图(PRM)的方法。该方法借鉴了基于采样的运动规划方法的思想,并结合多种组合优化技术,将原本连续状态空间中的轨迹规划问题转化为离散拓扑图上的路由问题。求解过程分为离线预处理和在线查询两个阶段。离线阶段采用Halton拟随机采样算法及Noon-Bean转换方法,将原问题转化为经典的非对称旅行商问题(ATSP);在线阶段根据战场态势的实时变化,快速更新路标图,然后采用LKH算法在线求解问题的近似最优解。为了保证生成的飞行轨迹满足平台的运动学/动力学约束,算法基于Gauss伪谱法构建了局部轨迹规划器。最后,以攻击时间最短为优化指标对算法进行了仿真实验。结果表明,本文提出的方法能够以较高的精度和在线收敛速度生成真实可行的、较优的多目标攻击轨迹。 The three-dimensional trajectory planning problem of single ground unmanned combat aircraft (UCAV) attacking multiple ground targets is studied. First, the problem is formalized as a special type of traveling salesman problem (TSP), a neighborhood-driven TSP with dynamics constraints (DCTSPN). Secondly, aiming at the problem of over-dimensioned planning space and excessive search cost, a method based on probabilistic roadmap (PRM) is proposed. The method draws on the idea of ​​sample-based motion planning method and combines a variety of combinatorial optimization techniques to transform the trajectory planning problem in the original continuous state space into the routing problem on the discrete topological graph. The solution process is divided into two stages: offline preprocessing and online inquiry. In the off-line phase, the Halton quasi-random sampling algorithm and the Noon-Bean conversion method are used to transform the original problem into the classical asymmetric traveling salesman problem (ATSP). In the online phase, the roadmap is rapidly updated according to the real- Approximate Optimal Solution to Problem. In order to ensure that the generated flight trajectory satisfies the kinematic / kinematic constraints of the platform, the algorithm builds a local trajectory planner based on Gauss pseudospectral method. Finally, the algorithm is simulated by taking the attack time as the optimization index. The results show that the proposed method can generate realistic and better multi-target attack trajectories with higher accuracy and on-line convergence speed.