Scholarlink

<p>This paper proposes a learning-based tracking strategy for non-cooperative spacecraft, taking into account limited onboard computational resources as well as constraints on relative state and control input. Initially, a barrier function is employed to map the relative state and the control inputs of the pursuer into a predefined region. Subsequently, an Integral Reinforcement Learning-based Hamilton-Jacobi-Isaacs equation is developed, which decreases dependence on system dynamics. To conserve computational efficiency, we first develop a static event-triggered control framework and then extend it to a dynamic formulation for enhanced resource conservation. Finally, a critic-only neural network is employed to approximate the value function and control strategy, with a meta-learning framework enabling rapid adaptation to new scenarios through a learned general initialization. The uniform ultimate boundedness of the state and the weight estimation error are proven by Lyapunov theory. The efficacy of the proposed approach is validated through comprehensive numerical simulations.</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

Learning-based event-triggered tracking strategy for non-cooperative spacecraft

  • 期刊论文

  • 发表日期：

    2025年9月4日

查看全文
  • Yu Shan,
  • Aixue Wang,
  • Guan Wang,
  • Hongwei Xia,
  • Guangcheng Ma

摘要

This paper proposes a learning-based tracking strategy for non-cooperative spacecraft, taking into account limited onboard computational resources as well as constraints on relative state and control input. Initially, a barrier function is employed to map the relative state and the control inputs of the pursuer into a predefined region. Subsequently, an Integral Reinforcement Learning-based Hamilton-Jacobi-Isaacs equation is developed, which decreases dependence on system dynamics. To conserve computational efficiency, we first develop a static event-triggered control framework and then extend it to a dynamic formulation for enhanced resource conservation. Finally, a critic-only neural network is employed to approximate the value function and control strategy, with a meta-learning framework enabling rapid adaptation to new scenarios through a learned general initialization. The uniform ultimate boundedness of the state and the weight estimation error are proven by Lyapunov theory. The efficacy of the proposed approach is validated through comprehensive numerical simulations.