Scholarlink

This paper investigates the active defense guidance problem for hypersonic vehicles in a target-interceptor-defender scenario. In this engagement, a target hypersonic vehicle with active defense attempts to evade the interceptor. Considering practical application scenarios, there always exist the observation noise, incomplete observation information, and uncertain interceptor’s guidance strategies simultaneously for the target hypersonic vehicle. These noises and uncertainties need to be handled by the active defense guidance. By regarding the active defense scenario as a partially observable Markov decision process, this paper proposes a novel guidance method with an adaptive dueling double deep Q-network (AD3QN) reinforcement learning algorithm. The AD3QN algorithm leverages a novel network fusion structure comprising a convolutional neural network (CNN) and a gated recurrent unit (GRU) network. Moreover, an incomplete information processing mechanism is proposed to deal with the noise and uncertainties in practical application scenarios. By utilizing CNN to extract features from the stacked data and integrating GRU, the AD3QN-based guidance enhances its ability for estimation and identification with incomplete time series information. Finally, numerical experiments are introduced to demonstrate the effectiveness and performance of the proposed active defense guidance.

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

Adaptive Active Defense Guidance for Hypersonic Vehicle with Incomplete Information Based on Reinforcement Learning

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Weilin Ni,
  • Peihuan Qiu,
  • Haizhao Liang

摘要

This paper investigates the active defense guidance problem for hypersonic vehicles in a target-interceptor-defender scenario. In this engagement, a target hypersonic vehicle with active defense attempts to evade the interceptor. Considering practical application scenarios, there always exist the observation noise, incomplete observation information, and uncertain interceptor’s guidance strategies simultaneously for the target hypersonic vehicle. These noises and uncertainties need to be handled by the active defense guidance. By regarding the active defense scenario as a partially observable Markov decision process, this paper proposes a novel guidance method with an adaptive dueling double deep Q-network (AD3QN) reinforcement learning algorithm. The AD3QN algorithm leverages a novel network fusion structure comprising a convolutional neural network (CNN) and a gated recurrent unit (GRU) network. Moreover, an incomplete information processing mechanism is proposed to deal with the noise and uncertainties in practical application scenarios. By utilizing CNN to extract features from the stacked data and integrating GRU, the AD3QN-based guidance enhances its ability for estimation and identification with incomplete time series information. Finally, numerical experiments are introduced to demonstrate the effectiveness and performance of the proposed active defense guidance.