An Algorithm for Robust Filtering of Angular Motion Parameters of Navigation Spacecraft

Abstract

The reliable performance of satellite constellations depends on onboard attitude control systems that deliver the required precision. Although Sun and star sensors are widely used, their known shortcomingsȔhigh cost, a narrow field of view (which causes navigation failures during angular motion), and heavy computational loadȔhave motivated the development of attitude estimation methods that rely solely on magnetometer measurements, even during agile manoeuvres. The narrow field of view often results in a loss of lock during fast reorientations, making magnetometers a fallback solution. The main difficulty lies in the interference that contaminates magnetometer readings due to various random factors. The extended Kalman filter (EKF) is commonly employed to mitigate this interference. However, existing EKF-based approaches assume specific motion regimes, use oversimplified noise models, and require prohibitive computational resources for onboard navigation spacecraft (NSC) computers. Moreover, they tend to fail when the actual noise statistics deviate from the assumed Gaussian properties, which limits their practical applicability. To overcome these issues, this paper presents a novel robust attitude estimation algorithm for NSC. The algorithm is invariant to the angular motion and guarantees stability and accuracy even when the magnetometer noise statistics are completely unknown. Its synthesis constructs the right-hand side of the estimation equation by minimising a functional defined over the least favourable class of noise distributions, thereby ensuring optimal performance in the worst-case scenario. In contrast to the EKF, the proposed estimator does not require tuning of noise covariance matrices and avoids linearisation errors. In addition to its robustness, the estimator has a drastically reduced dimensionȔequal only to the dimension of the estimated angular parameter vectorȔwhich significantly lowers the computational cost compared with the EKF and makes it suitable for onboard implementation. Finally, the results of numerical experiments are presented, illustrating the effectiveness and performance of the proposed approach under dynamic conditions.