Center-of-Rotation-Augmented Kalman Filter for Adaptive Attitude Reference System

This paper proposes a novel augmented Kalman filter-based attitude reference system (ARS) that uses an inertial sensor comprised of a tri-axial gyroscope and a tri-axial accelerometer. For accurate estimation of attitude using an inertial sensor, effective compensation of the non-gravitational acceleration is crucial. The proposed method resolves this issue by using a novel rotational motion detector to adaptively eliminate non-gravitational acceleration. The types of motions that the system experiences are accurately distinguished by augmenting center of rotation to the state vector. Due to our unconventional augmented state vector, the reformed filter properties have been thoroughly examined, and an observability analysis has been carried out. An extensive experimental validation was conducted under six diversified scenarios from the author-collected and open-source datasets, including both rotation-only and translation-rotationcombined motions. The results demonstrate that the proposed method accurately estimates attitude with sub-degree errors for most trials, proving robustness and accuracy under various motions. A comparative analysis reveals that our method outperforms the conventional method and the MTx algorithm.

Automated summary

This paper proposes a novel augmented Kalman filter-based attitude reference system (ARS) that uses an inertial sensor comprised of a tri-axial gyroscope and a tri-axial accelerometer. The proposed method effectively compensates for non-gravitational acceleration and accurately estimates attitude by adaptively eliminating non-gravitational acceleration using a rotational motion detector. Experimental validation conducted under diversified scenarios demonstrates the robustness and accuracy of the proposed method, outperforming conventional methods and the MTx algorithm.