无人机集群由于作业效率高、抗毁重构性强、能执行复杂危险任务，在军民领域得到广泛应用。无人机集群航迹规划是无人机协同完成任务的基础和关键，针对无人机集群航迹规划开展研究具有重要意义。目前，对于城市街道这类障碍物较密集、未知障碍物较多的复杂场景，无人机集群协同航迹规划的结果存在安全性低，效率低等问题。本文采用改进粒子群算法(Improved Particle Swarm Optimization，IPSO)和改进DQN(Improved Deep Q Network，IDQN)算法来实现静态和动态场景下的无人机集群航迹规划，旨在获得安全性高、平滑性好、耗能低的无人机航迹。具体的研究工作及成果如下：

(1)针对粒子群(Particle Swarm Optimization，PSO)算法规划航迹时存在的航迹冗余、安全性低的问题，提出了综合航程和安全性的IPSO算法。采用该算法规划无人机初始航迹，再进行航迹优化处理，仿真实验结果表明IPSO算法在密集障碍物场景下寻路速度快，冲突消解能力强，航迹安全性高。

(2)针对传统DQN算法网络训练时间长的问题，提出了分段DQN算法，该算法网络训练前期无人机步长是单调递减的，训练后期无人机步长固定不变。将分段DQN算法与Nature DQN、DDQN(Double Deep Q Network)、Dueling DQN (Dueling Deep Q Network)等算法进行对比，仿真实验结果显示分段DQN算法在寻路成功率、收敛速度、网络训练效率上具有较大优势。

(3)针对复杂动态场景，提出了分段DQN-IPSO融合航迹规划算法。该算法分为两个阶段：第一阶段由分段DQN算法规划全局航迹，第二阶段由IPSO算法规划局部航迹。并且提出了基于拍卖算法的任务分配方法解决任务分配问题。仿真实验结果表明本文提出的方法能指导无人机集群避开未知动态障碍物，安全到达目标点。

(4)基于Unity平台开发了一个可展示无人机飞行状态和轨迹的三维仿真环境，在该环境下验证了航迹规划算法的有效性和安全性，并证明了任务分配环节对无人机集群航迹规划的重要性。

Unmanned aerial vehicle(UAV) swarms are widely used in the military and civilian fields due to their high operating efficiency, strong survivability and reconfigurability, and the ability to perform complex and dangerous tasks. UAV swarm trajectory planning is the basis and key for UAVs to complete tasks cooperatively. It is of great significance to conduct research on UAV swarm trajectory planning. At present, for complex scenes such as urban streets with dense obstacles and many unknown obstacles, the results of collaborative trajectory planning of UAV clusters have problems such as low safety and low efficiency. In this paper, the Improved Particle Swarm Optimization(IPSO) and the Improved Deep Q Network(IDQN) algorithms are used to realize UAV cluster track planning in static and dynamic scenarios, aiming to obtain high security and smooth UAV track with good performance and low energy consumption. The specific research work and results are as follows:

(1)Aiming at the problems of track redundancy and low safety in particle swarm optimization(PSO) algorithm planning track, an IPSO algorithm integrating range and safety is proposed,The algorithm is used to plan the initial trajectory of the UAV, and then the flight path optimization is carried out. The experimental results show that the IPSO algorithm has fast pathfinding speed, strong conflict resolution ability and high track security in the dense obstacle scene.

(2)Aiming at the problem of long network training time of the traditional DQN algorithm, a segmented DQN algorithm is proposed. The step size of the UAV in the early stage of network training of the algorithm is monotonically decreasing, and the step size of the UAV in the later stage of training is fixed. Comparing the segmented DQN algorithm with Nature DQN, DDQN(Double Deep Q Network), Dueling DQN(Dueling Deep Q Network) and other algorithms, the results show that the segmented DQN algorithm has advantages in pathfinding success rate, convergence speed, and network training efficiency.

(3)For complex dynamic scenarios, a segmented DQN-IPSO fusion trajectory planning algorithm is proposed. The algorithm is divided into two stages: in the first stage, the segmented DQN algorithm is used to plan the global track, and in the second stage, the IPSO algorithm is used to plan the local track. And a task allocation method based on auction algorithm is proposed to solve the task allocation problem. The experimental results show that the method proposed in this paper can guide the UAV swarm to avoid unknown dynamic obstacles and reach the target point safely.

(4) Based on the Unity platform, a 3D simulation environment that can display the flight status and trajectory of UAVs was developed. In this environment, the validity and safety of the trajectory planning method were verified, and it was proved that the task assignment link has a great impact on UAV swarm navigation.