摘要
为了提高无人作战效率,将深度强化学习应用于雷达对抗侦察无人机的航线规划中.首先对雷达对抗的侦察行动进行分析,建立无人机飞行航线仿真与分幕评价侦察效果的模型;然后根据训练强化学习智能体的需求,对规划航线过程中所需信息进行了参数化,以雷达对抗侦察行动的特点设计了动作与奖励机制,给出了适用的智能体神经网络结构和训练算法.仿真结果表明,与两种固定航线的规划策略相比,采用本文深度强化学习的航线规划方法,在平均收益上提高了37%,完成既定侦察任务次数提高了超过1.5倍,完成任务平均消耗的航程减少了13%.
In order to improve the efficiency of unmanned combat,deep reinforcement learning is applied to the route planning of radar countermeasure reconnaissance UAV.Firstly,this paper analyzes radar countermeasure reconnaissance operation,and establishes the model of UAVflight path simulation and split-screen evaluation of reconnaissance effect.Then,according to the needs to train reinforcement learning intelligent agent,the paper parameterizes the required information in the course of route planning,and finally,accords to the characteristics of radar countermeasure reconnaissance to design the actions and reward mechanism,presenting the applicable structure of the intelligent agent neural network and the training algorithm.Simulation results show that compared with the two fixed route planning strategies,after using the route planning method of deep reinforcement learning in this paper,the average return has been increased by 37%,the number of completing the set reconnaissance missions increased by more than 1.5 times,and the cruising range for completing the missions reduced by 13% on average.
作者
高石印
石玮
王聪阜
刘辉
GAO Shiyin;SHIWei;WANG Congfu;LIU Hui(No.93046 Unit,the PLA,Qingdao 266109,China)
出处
《空天预警研究学报》
CSCD
2023年第2期119-123,共5页
JOURNAL OF AIR & SPACE EARLY WARNING RESEARCH
关键词
雷达对抗
无人机侦察
航线规划
深度强化学习
radar countermeasure
UAV reconnaissance
route planning
deep reinforcement learning
