中国空间科学技术

中国空间科学技术 ›› 2025, Vol. 45 ›› Issue (6): 133-140.doi: 10.16708/j.cnki.1000-758X.2025.0095

• 论文 • 上一篇    下一篇

基于可观测度的自主导航序列特征轻量构建方法

李嘉兴1，王大轶1,*，徐超2，李茂登2，董天舒1，符方舟3   

  1. 1.北京空间飞行器总体设计部，北京100094
    2.北京控制工程研究所，北京100094
    3.中山大学 航空航天学院，深圳518107
  • 收稿日期:2024-08-12 修回日期:2024-12-02 录用日期:2025-01-04 发布日期:2025-11-17 出版日期:2025-12-01

Observability degrees based lightweight construction method for sequential features of autonomous navigation

LI Jiaxing1，WANG Dayi1,*，XU Chao2，LI Maodeng2，DONG Tianshu1，FU Fangzhou3   

  1. 1.Beijing Institute of Spacecraft System Engineering，Beijing 100094，China
    2.Beijing Institute of Control Engineering，Beijing 100094，China
    3.School of Aeronautics and Astronautics，Sun Yat-sen University，Shenzhen 518107，China
  • Received:2024-08-12 Revision received:2024-12-02 Accepted:2025-01-04 Online:2025-11-17 Published:2025-12-01

PDF (PC)

0

摘要： 基于序列图像的行星着陆自主导航受到着陆器上计算资源限制，难以处理大量图像特征、难以匹配大量图像。从空间和时间两个维度设计可观测度指标，陆标空间构型指标能够优化序列特征中的最佳特征位置；深度估计误差指标能够优化序列特征的最佳间隔时间。用于指导构建计算负担小的轻量化序列特征。提出的可观测度指标不需复杂的矩阵运算，可适用于着陆器上自主优选陆标和特征序列。仿真结果表明，相比于传统方法，提出的序列特征轻量构建方法能显著降低陆标选择时的计算量和图像处理次数，有效提升着陆器的自主导航能力。

关键词: 行星着陆, 自主导航, 序列图像, 可观测度, 轻量化

Abstract: Autonomous planetary landing navigation based on sequential images is constrained by the limited computational resources onboard the lander, making it difficult to process and match a large number of image features. This paper designs observability degrees from both spatial and temporal dimensions. A spatial configuration observability degree for landmarks is introduced to optimize the selection of the best feature locations within the sequence, while a depth estimation error observability degree is proposed to determine the optimal time intervals between features. These observability degrees are used to guide the construction of lightweight sequential features with low computational burden. The proposed observability degrees avoid complex matrix operations, making them suitable for onboard autonomous selection of landmarks and feature sequences. Simulation results demonstrate that, compared to traditional methods, the lightweight sequential feature construction method significantly reduces computational load and the number of image processing operations during landmark selection, thereby effectively enhancing the autonomous navigation capability of the lander.

Key words: planetary landing, autonomous navigation, sequential image, observability degree, lightweight