We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
Adaptive invariant Kalman filtering for lie groups attitude estimation with biased and heavy-tailed process noise.
- Authors
Wang, Jiaolong; Zhang, Chengxi; Liu, Jinyu; Wei, Caisheng; Liu, Haitao
- Abstract
Attitude determination is fundamental for spacecraft missions in aerospace engineering. Kalman filter (KF) is the optimal estimator in least square sense and, using the symmetry properties of matrix Lie groups system, the invariant Kalman filter (IKF) has been developed to boost the filtering performance for attitude estimation. However, due to presence of frequent and severer maneuvers, the Lie groups attitude dynamics is usually corrupted by significant biases and heavy-tailed outliers, which usually leads to decreased precision of IKF. For attitude estimation problem troubled by biased and heavy-tailed process noise, this work proposes a new invariant Kalman filter (VBAIKF) by constructing the hierarchical Gaussian system model: the probability density function of prior estimate state is first described using the student's t distribution, while the unknown scale covariance matrix and degrees of freedom (dof) of the employed student's t distribution are defined as the inverse Wishart distribution (IWD) and Gamma distribution. In VBAIKF, the Lie groups rotation matrix of spacecraft, the biased mean, the parameters of dof and scale covariance matrix are online estimated together by variational Bayesian fixed-point iterations. The simulation results with Lie groups attitude estimation system further verify the superior filtering adaptability and precision of proposed approach VBAIKF than other methods for attitude determination with biased mean and heavy-tailed process noise.
- Subjects
KALMAN filtering; GAUSSIAN processes; PROBABILITY density function; GAMMA distributions; AEROSPACE engineering; COVARIANCE matrices; LEAST squares
- Publication
Transactions of the Institute of Measurement & Control, 2023, Vol 45, Issue 2, p249
- ISSN
0142-3312
- Publication type
Article
- DOI
10.1177/01423312221110956