基于改进DQN的直升机CGF突防路径规划  

Helicopter CGF Penetration Path Planning Based on Improved DQN

作　　者：许强强 李克奇 岳忠奇 杨艳良 岳晋忠 XU Qiangqiang;LI Keqi;YUE Zhongqi;YANG Yanliang;YUE Jinzhong(North Automatic Control Technology Institute,Taiyuan 030006,China)

机构地区：[1]北方自动控制技术研究所,太原030006

出　　处：《火力与指挥控制》2025年第1期104-112,共9页Fire Control & Command Control

摘　　要：针对直升机CGF突防雷达阵地时的路径规划问题,提出了一种基于改进DQN的直升机CGF的突防路径规划方法。结合了人工势场原理和专家经验,对传统的深度强化学习DQN算法进行了多方面的改进,以适应动态和不可预知的战场环境。通过在MetaSim仿真作战平台中进行实验,证明该方法在路径规划的平均所需时间上比传统A*和DIJKSTRA算法以及传统DQN算法有所提高,同时在路径长度和成功率上也显示出显著的改进。In order to solve the problem of path planning of helicopter CGF penetration radar posi-tion,a method of path planning of helicopter CGF penetration based on improved DQN is proposed.The artificial potential field principle and expert experience are combined,the traditional deep reinforcement learning DQN algorithm is improved in many aspects to adapt to the dynamic and unpredictable battle-field environment.The experiments on MetaSim simulation combat platform are carried out and show that the proposed method improves the average time required for path planning compared with traditional A*,DIJKSTRA and DQN algorithms,and also shows the significant improvements in path length and success rate.

关 键 词：路径规划 计算机生成兵力 深度强化学习 人工势场 DQN 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]