부산대학교 도서관

In this study, a pioneering approach is introduced to elevate the performance of angular rate estimation in scenarios where gyros are not available. Within situations involving angular velocities, star observations exhibit notable variations in centroid accuracy based on their brightness. The critical aspect is the dispersion of starlight across multiple pixels due to the movement of the sensor during the exposure time. These effects culminate in the blurring of star intensity, resulting in the degradation of centroid accuracy. This phenomenon is particularly accentuated in dim stars with a low signal-to-noise ratio. Therefore, to enhance the performance of angular rate estimation, it is imperative to integrate these accuracy differences into the estimation process. The proposed algorithm originates from the idea of detecting faults in parity space, aiming at fault detection and isolation (FDI). Stars observed by the star sensor are considered redundant to each other and are used to identify stars with unusually large errors. It is observed that when the proposed method is applied, there is an approximate 40% reduction in the average root mean square error (RMSE) of the estimated angular velocity compared with when quality is not considered at all. In addition, an improvement of around 9% is noted compared with simply excluding values detected as faults.