系统工程与电子技术 ›› 2024, Vol. 46 ›› Issue (12): 4183-4191.doi: 10.12305/j.issn.1001-506X.2024.12.27

• 制导、导航与控制 • 上一篇    

基于GPR-UKF的天文测角/测速组合导航方法

张寿健1,2, 桂明臻1,2,*   

  1. 1. 中南大学自动化学院, 湖南 长沙 410083
    2. 智慧地球重点实验室, 北京 100029
  • 收稿日期:2024-01-30 出版日期:2024-11-25 发布日期:2024-12-30
  • 通讯作者: 桂明臻
  • 作者简介:张寿健 (2002—), 男, 硕士研究生, 主要研究方向为天文导航与机器学习
    桂明臻 (1992—), 男, 副教授, 博士, 主要研究方向为航天器自主导航方法
  • 基金资助:
    智慧地球重点实验室基金(KF2023ZD01-01)

Navigation method using angle/velocity measurement based on GPR-UKF

Shoujian ZHANG1,2, Mingzhen GUI1,2,*   

  1. 1. School of Automation, Central South University, Changsha 410083, China
    2. Key Laboratory of Smart Earth, Beijing 100029, China
  • Received:2024-01-30 Online:2024-11-25 Published:2024-12-30
  • Contact: Mingzhen GUI

摘要:

在以太阳作为目标源的天文测速导航中,多普勒速度量测量存在较多野值误差,严重影响导航精度。对此,提出一种基于高斯过程回归与无迹卡尔曼滤波(Gaussian process regression and unscented Kalman filtering,GPR-UKF)的野值检测修复方法,建立速度量测量的动态预测模型。此外,还针对不同参数对模型精度的影响进行研究。经仿真验证,所提方法效果显著优于传统野值处理方法。

关键词: 组合导航, 高斯过程回归, 无迹卡尔曼滤波, 太阳多普勒速度, 星光角距, 野值处理

Abstract:

In astronomical velocity measurement navigation with the Sun as the target source, there are many outliers in Doppler velocity measurement, which seriously affects the accuracy of navigation. Thus, a outlier detection and repair method based on Gaussian process regression and unscented Kalman filtering (GPR-UKF) is proposed to establish a dynamic prediction model for velocity measurement. In addition, the impact of different parameters on the accuracy of the model is researched. Simulation verification test demonstrates that the proposed method has better performance than traditional outlier processing methods.

Key words: integrated navigation, Gaussian process regression (GPR), unscented Kalman filtering (UKF), solar Doppler velocity, starlight angle pitch, outlier processing

中图分类号: