Attitude filtering with uncertain process and measurement noise covariance using SVD-aided adaptive UKF

Abstract

It is presented in this article how to simultaneously alter the process and measurement noise covariance matrices for nontraditional attitude filtering technique. The unscented Kalman filter (UKF) and singular value decomposition (SVD) methods are integrated in the nontraditional attitude filtering algorithm to estimate a nanosatellite's attitude with an inherent robustness feature. The SVD approach determines the attitude of the nanosatellite and provides one estimate at a single frame utilizing measurements from the magnetometer and Sun sensor as the initial stage of the algorithm. These attitude terms are subsequently fed into the UKF with their error covariances, which makes the filter robust inherently. The attitude estimations of the satellite are compared between the filters presented. The Q (process noise covariance) adaption approach with multiple scale factors is specifically suggested for differences in between the process channels. Performance of the multiple scale factors-based adaptive SVD-aided UKF is examined in the event of process noise increase, which may result from changes in the environment or satellite dynamics.