Trajectory planning is crucial for Unmanned Aerial Vehicles (UAVs) applications. It is still a challenge to plan UAV trajectories in large-scale complicated three-dimensional urban environments, mainly due to the complexities associated with handling a tremendous number of obstacles. To relieve this difficulty, this study proposes a two-stage control network approach. The first stage is an offline planning stage, aiming at constructing a discrete control network in the working environment achieving collision avoidance and meanwhile considering UAV kinetic properties. The second stage is an online planning stage, where a feasible corridor is generated based on the shortest path in the control network, and the obtained route is smoothed by a three-dimensional Dubins curve algorithm; then, a time-optimal trajectory generation algorithm is applied to trace the designed spatial path. Numerical studies based on a virtual city as well as a large-scale real-world urban area are conducted. The results show that the proposed approach is able to identify reasonable UAV trajectories within negligible computational time.

轨迹规划对于无人飞行器（UAV）应用至关重要。在大型复杂的三维城市环境中规划无人机轨迹仍然是一项挑战，主要是由于与处理大量障碍相关的复杂性。为了缓解这一困难，本研究提出了一种两阶段控制网络方法。第一个阶段是离线计划阶段，旨在在工作环境中构建离散控制网络，从而避免碰撞并同时考虑无人机的动力学特性。第二阶段是在线规划阶段，在该阶段基于控制网络中的最短路径生成一条可行的走廊，并通过三维杜宾斯曲线算法对获得的路线进行平滑处理；然后，应用时间最优轨迹生成算法来跟踪设计的空间路径。进行了基于虚拟城市以及大规模现实世界城市区域的数值研究。结果表明，该方法能够在可忽略的计算时间内识别出合理的无人机航迹。