Scholarlink

<p>This paper proposes an integrated guidance and evasion method for surface-to-air missiles engaging a maneuvering target capable of launching defender missiles. The method is developed within a hierarchical reinforcement learning framework. Two low-level agents are trained independently using a meta-reinforcement learning approach with recurrent neural networks to execute the skills of guiding the missile toward the target and evading the defender missile. At the high level, a discrete action policy is combined with a threat-region-based switching mechanism to select and maintain an appropriate option over an extended time interval. The proposed method is evaluated across multiple scenarios and compared with several conventional methods. The results show that the proposed method achieves high guidance accuracy, limits control energy and flight time, and demonstrates robust adaptability across diverse engagement scenarios.</p>

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

Reinforcement Learning-Based Integrated Guidance and Evasion for Surface-to-Air Missiles Against Defended Targets

  • 期刊论文

  • 发表日期：

    2026年6月24日

查看全文
  • Tran Quang Minh,
  • Cao Huu Tinh

摘要

This paper proposes an integrated guidance and evasion method for surface-to-air missiles engaging a maneuvering target capable of launching defender missiles. The method is developed within a hierarchical reinforcement learning framework. Two low-level agents are trained independently using a meta-reinforcement learning approach with recurrent neural networks to execute the skills of guiding the missile toward the target and evading the defender missile. At the high level, a discrete action policy is combined with a threat-region-based switching mechanism to select and maintain an appropriate option over an extended time interval. The proposed method is evaluated across multiple scenarios and compared with several conventional methods. The results show that the proposed method achieves high guidance accuracy, limits control energy and flight time, and demonstrates robust adaptability across diverse engagement scenarios.