Observability analysis and design of two nested filters for the satellite attitude estimation with magnetometer‐only

Abstract

This study aims to propose a scenario for determining the attitude of a nano-satellite with only a magnetometer in the case of temporal failure or faults in other sensors of attitude determination subsystem. For the sake of this matter, two nested filters are employed. The first filter estimates the magnetic field derivative by an extended Kalman filter (EKF). This information is applied along with the magnetometer outputs to obtain the attitude in the second one. A proposed algorithm based on the Square-Root UnScented QUaternion Estimator has been carried out to guarantee the positiveness of the covariance matrices considering the restriction about the unit norm of the quaternion. Furthermore, the non-linear observability of these filters is investigated analytically accompanied by condition number studies to show the functionality of these filters for magnetometer-only attitude determination. Finally, Monte Carlo simulation results indicate better performance of this approach in comparison with the traditional method of multiplicative EKF, particularly with large initialisation errors.