摘要
传统的制导规律存在信息依赖度高、对目标机动样式适应能力不足等问题。针对空空导弹攻击机动目标作战使用场景,基于深度强化学习理论,构建适应于空中机动目标制导的智能学习场景,提出基于深度强化学习的系数时变最优制导律,并采用改进的PPO算法,完成了制导参数实时调节神经网络的训练及部署,最后通过数学仿真验证了优化策略的正确性。
Traditional guidance laws have problems such as high information dependence and insufficient adaptability to target maneuvering styles.Based on the theory of deep reinforcement learning,an intelligent learning scenario suitable for air maneuvering target guidance is constructed for the combat scenario of air-to-air missile attack on moving targets.A coefficient time-varying optimal guidance law based on deep reinforcement learning is proposed,and an improved PPO algorithm was adopted to complete the training and deployment of a neural network for real-time adjustment of guidance parameters.Finally,the correctness of the optimization strategy is verified through digital simulation.
作者
周桃品
宋丹阳
龚铮
ZHOU Tao-pin;SONG Dan-yang;GONG Zheng(China Airborne Missile Academy,Luoyang,Henan 471009,China;Nation Key Laboratory of Air-based Information Perception and Fusion,Luoyang,Henan 471009,China)
出处
《电子技术与软件工程》
2025年第2期12-18,共7页
Electronic Technology & Software Engineering
基金
航空科学基金(No.2020Z037012003,No.2023Z037012002)
关键词
智能制导
深度强化学习
最优制导律
神经网络
近端策略优化
intelligent guidance
deep reinforcement learning
optimal guidance law
neural network
proximal policy optimization algorithm
