Sensor alignment estimation and calibration for a LEO satellite with validation by on-orbit data

Abstract

The Multiplicative Extended Kalman Filter (MEKF) is a commonly used algorithm for satellite attitude estimation that estimates attitude errors and changes in bias. This filter utilizes star tracker absolute attitude and gyroscope angular rate measurements and fuses them to reduce noise in sensor measurements. However, real-world environments and launch effects introduce additional error terms that require further error modeling and estimation. An MEKF with augmented states can be used to estimate these quantities, accounting for gyroscope misalignment and scale factor errors. The gyroscope is selected for calibration since it can be calibrated relative to a highly accurate attitude sensor, such as a star tracker, enabling estimation accuracy comparable to that of the star tracker. It is crucial to distinguish scale factor and misalignment errors from bias errors, as these errors tend to appear constant and unobservable when the satellite is not undergoing maneuvers. To effectively address this, a maneuver-based test was conducted using the LEO imaging satellite platform, which provided the necessary excitation for accurate calibration. The estimation algorithm was developed accordingly. Initially, the algorithm was validated in a controlled simulation environment designed to replicate in-orbit conditions, followed by application to real in-orbit data to estimate the relevant error states. Various maneuvers were analyzed in simulation to evaluate observability and the optimal maneuver. By considering both observability and the maneuvering capability of the LEO imaging platform, a polynomial maneuver combining rotation in all three axes was selected for the calibration. The results demonstrated that the assumed misalignment and scale factor errors have a significant impact on the overall estimation accuracy.