Scholarlink

<p>In space situational awareness (SSA), non-cooperative targets are monitored under complex illumination and dynamic conditions, where reliable quantitative observation evaluation is essential for threat detection and catalog maintenance. This paper addresses the challenge of constructing a space-based observation evaluation framework for non-cooperative targets under complex illumination and dynamic conditions, and develops an artificial neural network (ANN) algorithm for optimizing comprehensive observation effectiveness-evaluation parameters within a typical rendezvous and proximity operations (RPO) dynamics framework. The study integrates constraints such as solar incidence angle, field of view, and relative angular rate, forming a multi-constraint finite-horizon set. A comprehensive effectiveness-evaluation model is proposed, incorporating submodels for relative distance, image motion, effective observation time, and coverage, with dynamic weighting based on mission priority. The comprehensive observation effectiveness evaluation problem is formulated as a highly nonlinear optimization task, and an improved reinforcement learning neural network algorithm with feedback coefficient scheduling (FCS-RLNNA) is introduced. Numerical simulations in GEO scenarios demonstrate that FCS-RLNNA outperforms existing optimization methods in terms of convergence speed and solution stability. The proposed framework effectively enhances multi-constraint RPO observation while satisfying safety and imaging-quality requirements.</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

Multi-constraint spacecraft close-range observation effectiveness evaluation and optimization via a feedback-coefficient scheduled RLNNA

  • 期刊论文

  • 发表日期：

    2026年5月23日

查看全文
  • Hongsheng Hu,
  • Yufei Luo,
  • Yunhe Meng

摘要

In space situational awareness (SSA), non-cooperative targets are monitored under complex illumination and dynamic conditions, where reliable quantitative observation evaluation is essential for threat detection and catalog maintenance. This paper addresses the challenge of constructing a space-based observation evaluation framework for non-cooperative targets under complex illumination and dynamic conditions, and develops an artificial neural network (ANN) algorithm for optimizing comprehensive observation effectiveness-evaluation parameters within a typical rendezvous and proximity operations (RPO) dynamics framework. The study integrates constraints such as solar incidence angle, field of view, and relative angular rate, forming a multi-constraint finite-horizon set. A comprehensive effectiveness-evaluation model is proposed, incorporating submodels for relative distance, image motion, effective observation time, and coverage, with dynamic weighting based on mission priority. The comprehensive observation effectiveness evaluation problem is formulated as a highly nonlinear optimization task, and an improved reinforcement learning neural network algorithm with feedback coefficient scheduling (FCS-RLNNA) is introduced. Numerical simulations in GEO scenarios demonstrate that FCS-RLNNA outperforms existing optimization methods in terms of convergence speed and solution stability. The proposed framework effectively enhances multi-constraint RPO observation while satisfying safety and imaging-quality requirements.