Scholarlink

In order to attack maneuvering targets, high gains (i.e. navigation gain) are required in the proportional navigation guidance (PNG) laws, by which the commands of overload change fast. Since the response delay of the autopilot cannot be ignored, it is difficult for the missile to track such a fast-varying command accurately with traditional guidance law. A deep reinforcement learning-based solution for PNG is proposed, by which the trade-off issue associated with the high-gain guidance law is addressed. In this solution, the guidance gain is adjusted adaptively according to the pre-trained policy network. By applying the solution to a simulation case with head-on target, the validity of the proposed guidance law is proved.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

An Adaptive Terminal Guidance Law Based on Deep Reinforcement Learning

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Qixin Dai,
  • Baoqing Yang,
  • Shiyao Li

摘要

In order to attack maneuvering targets, high gains (i.e. navigation gain) are required in the proportional navigation guidance (PNG) laws, by which the commands of overload change fast. Since the response delay of the autopilot cannot be ignored, it is difficult for the missile to track such a fast-varying command accurately with traditional guidance law. A deep reinforcement learning-based solution for PNG is proposed, by which the trade-off issue associated with the high-gain guidance law is addressed. In this solution, the guidance gain is adjusted adaptively according to the pre-trained policy network. By applying the solution to a simulation case with head-on target, the validity of the proposed guidance law is proved.