An Adaptive Attitude Estimation Method Based on SigKAN-Assisted Filtering
IEEE Sensors Journal · 01.07.2026 · 26 (13) · 20508-20517
DOI: 10.1109/jsen.2026.3697808 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Sensors Journal, erschienen am 01.07.2026. Verfasst von Peng Li und 2 weiteren Autor:innen. Die Arbeit nennt 42 Literaturangaben. Gefördert von: National Natural Science Foundation of China.
Abstract (Originalsprache)
Accurate 3-D attitude estimation using inertial measurement units (IMUs) remains challenging under dynamic motion and magnetic disturbance, primarily due to the integration drift of gyroscopes and the contamination of accelerometer and magnetometer measurements by linear acceleration and ferromagnetic disturbance. To address these issues, this article proposes a novel adaptive attitude estimation framework that synergistically integrates a Kalman filter with a signature-enhanced Kolmogorov–Arnold network (SigKAN) for robust and precise orientation tracking. First, a Kalman filter is designed to explicitly model and separate linear acceleration and magnetic disturbance by extending the state vector to include gravity, linear acceleration, geomagnetic field, and ferromagnetic disturbance. To mitigate modeling uncertainties arising from the state space and time-varying noise characteristics, a SigKAN-based adaptive strategy is introduced to dynamically learn and adjust the process noise covariance matrix in real time. The SigKAN network leverages path signatures to capture higher order temporal features of sensor data and employs the Kolmogorov–Arnold representation for interpretable nonlinear mapping, thereby enhancing noise modeling accuracy under non-Gaussian and nonstationary conditions. Finally, the refined gravity and geomagnetic estimates are fed into the quaternion estimator (QUEST) algorithm to compute the optimal attitude quaternion. The effectiveness of the proposed method is validated through experiments under four representative scenarios.
Förderung
- National Natural Science Foundation of China
Zitieren
APA
Li, P., Ni, W., & Zhang, W. (2026). An Adaptive Attitude Estimation Method Based on SigKAN-Assisted Filtering. IEEE Sensors Journal, 26(13), 20508-20517. https://doi.org/10.1109/jsen.2026.3697808
Vancouver
Li P, Ni W, Zhang W. An Adaptive Attitude Estimation Method Based on SigKAN-Assisted Filtering. IEEE Sensors Journal. 2026;26(13):20508-20517. doi:10.1109/jsen.2026.3697808
BibTeX
@article{Li2026,
author = {Li, Peng and Ni, Weilong and Zhang, Wen-An},
title = {An Adaptive Attitude Estimation Method Based on SigKAN-Assisted Filtering},
journal = {IEEE Sensors Journal},
year = {2026},
volume = {26},
number = {13},
pages = {20508--20517},
doi = {10.1109/jsen.2026.3697808}
}
EUVIMED