Autonomous Attitude Filtering for Terrestrial Star Sensors Under Maneuvering Conditions

In this article, a generalized maneuvering attitude error-state Kalman filter (GMAESKF) is proposed to mitigate the attitude measurement noise and fulfill the requirement of angular velocity estimation for terrestrial star sensors when they are applied to agile maneuvering carriers. Based on the interchangeability between quaternions and rotation matrices, this method defines a generalized representation and operations for attitude, as well as attitude dynamics, and further establishes a pseudolinear maneuvering attitude kinematics based on the Jerk model with high-dimensional characteristics. The attitude measurement model of the multi-field-of-view (multi-FOV) star sensor based on directional statistics is refined through maximum likelihood estimation, thereby deriving a generalized measurement equation. Based on these, an error-state Kalman filter (ESKF) for attitude, angular velocity, angular acceleration, and angular jerk is constructed, realizing the autonomous estimation of the required parameters. In the numerical simulation, the proposed method reduced the attitude measurement error by 42.3%, with an rms error of 0.0171°/s for angular velocity estimation. Both numerical simulations and flight experiments consistently show that, compared to existing related algorithms, the proposed algorithm yields more accurate results under agile maneuvering conditions.