期刊文献+

基于增强学习的空空导弹智能精确制导律研究 被引量:3

Research of Precision Guidance Law Based on Q-Learning for Air-to-Air Missile
下载PDF
导出
摘要 根据现代战争的对抗格局,提出了空空导弹拦截高速大机动目标的智能制导律。这种制导律是采用基于Q—learning算法的。Q—learning的思想是直接优化一个可迭代计算的Q函数,并利用增强学习实现知识的自动获取,来扩展所能得到的知识资源。在Q—learning算法中,系统通过计算状态的值函数或者状态-动作对的值函数来控制导弹的飞行。根据环境的评价性回报函数来实现决策的优化。从而能够达到行为优化。这种制导规律只需要导弹和目标的位置、状态变量和法向过载的测量量。易于弹上实时实现,并且将这种制导律和传统制导相比较。结果表明:这种制导具有一定的智能行为。可以拦截大机动目标。这种智能制导方法有利于提高打击精度和载机的作战生存能力。 On the basis of the combat in modern wars, a new intelligent guidance method, which is used in air-to-air missile to intercept high-speed mobile goal is presented in this paper. This theory is based on the differential games theory and Q-learning algorithm. The Q-learning is a method that directly optimizes an iterative and calculable Q-function, and automatically acquires knowledge using a reinforcing learning to enlarge the available knowledge resource. The system computes state value function or state-action value function to control the action of missile. The system makes the optimal action basis on the appraising function of the environment. It requires only relative position, states variable and trajectory normal load factor measurement. This guidance law is easily realized for realtime on missile. We compare this guidance law with other traditional guidance law. The results show that the intelligent guidance law has intelligent action to intercept high-speed mobile goal. This intelligent guidance theory can improve the attacking precision and the survivability of the fighter aircrafts.
出处 《战术导弹控制技术》 2006年第4期19-22,76,共5页
关键词 微分对策 机器学习 Q-LEARNING BP神经元网络 精确制导 Differential Games, Machine Learning, Q-Learning, BP Neural Network, Precision Guidance.
  • 引文网络
  • 相关文献

同被引文献71

引证文献3

二级引证文献38

;
使用帮助 返回顶部