摘要
设计了一种基于并行粒子群和RL(reinforcement learning,RL)的无人机航路规划方法.首先,定义了引入转角角度约束的航路规划的代价模型,然后,提出了一种基于RL中的Q-Learning算法的初始航路规划方法,将其作为粒子的初始位置,将粒子群划分为多个种群进行并行寻优,得到所有种群局部最优解,计算其最优值获取全局最优解;最后设计了基于并行粒子群和RL的无人机航路规划算法.仿真实验表明文中方法能实现复杂威胁情况下UCAC航路规划,具有较高的规划效率,且与其他方法相比,具有收敛速度快和全局寻优能力强的优点,具有较强的可行性和实用性.
Path planning for UCAV（Unmanned Combat Aerial Vehicle）is a NP problem.The existing method has the defects of slow convergence rate and easily getting the local solution.Therefore,apath planning method based on parallel particle swarm and RL has been proposed.Firstly,the anchor constraint is introduced to the cost model of path planning for UCAV.Then,the initial path planning method based on RL and Q-Learning is proposed which is used as the initial position for particle.The particle swarm is divided to multiple populations to parallel search,the best solution in all the optimal solutions is used as the global optimal solution.Finally,the path planning algorithm for UCAV is designed base on parallel particle swarm and RL.The simulation result shows the method proposed in this paper can achieve the path planning for UACA on the condition of compound threats with the higher planning efficiency.Compared with the other methods,it has the slow convergence and global search ability with stronger feasibility and applicability.
出处
《西南师范大学学报（自然科学版）》
CAS
北大核心
2016年第3期31-36,共6页
Journal of Southwest China Normal University(Natural Science Edition)
基金
四川省科技厅项目(2015GZ0279)
