AIAA Guidance, Navigation, and Control (GNC) Conference, Boston, MA, United States Of America, 19 - 22 August 2013
A method for in-orbit estimation of time-varying residual magnetic moment is presented for small satellite applications. The Unscented Kalman Filter is used as the estimator. Unlike the existing studies in the literature, unexpected abrupt changes in the residual magnetic moment are also considered. Sudden shifts in the mean of the residual magnetic moment is detected by a low pass filter applied to the normalized innovation of the Unscented Kalman Filter. Then, by using a simple approach, which does not require an excessive extra computational load, the covariance of the filter is adapted to get better tracking without sacrificing the estimation accuracy. Compared to the existing methods that reinitialize the filter or increase the process noise covariance after each change detection, the proposed method does not need a priori information about the magnitude of the shift and assures both accurate estimation and good tracking performance for shifts with different magnitudes.