Real-time tuning separate-bias extended Kalman filter for attitude estimation

Abstract

This paper provides a new method for robust spacecraft attitude estimation in the presence of measurement biases. The proposed method is developed based on the separate-bias or two-state Kalman filter which was first introduced by Friedland. The separate-bias Kalman filter consists of two stages: the first stage, the "bias-free" filter, is based on the assumption that the bias is nonexistent; the second stage, the "bias" filter, is implemented to estimate bias vectors. The output of the first filter is then corrected with the output of the second filter. In this research, the authors propose a real-time tuning method for a parameter in the Kalman gain calculation process of the "bias" filter. The adaptive scale factor is optimized relying on the minimization of the cost function, which is calculated from the difference between the predicted and measurement values. The proposed filter has a faster convergence speed from large initial errors and an increased accuracy on unpredicted bias models than conventional methods. Moreover, to verify these advantages, the research also provides analyses and comparisons between the proposed method with conventional methods like the original separate-bias Kalman filter, unscented Kalman filter and extended Kalman filter in several numerical simulation scenarios for a microsatellite.