In-flight on-line calibration method for MEMS gyroscope based on adaptive unscented Kalman filter algorithm

The attitude sensors need to be calibrated on-line in order to guarantee the performance of system in the application of the micro-satellite. A real-time drift error estimation method of MEMS gyroscope is presented by using three-axis magnetometer measurements without any external attitude reference. The unscented Kalman filter (UKF) is applied as the optimal estimation algorithm, the gyroscope drift is used as the filter state vector, and the finite difference of magnetometers observation is established as the measurement vector. Since the measurements of the magnetometers are susceptible to interferences, and this would lead to inaccuracy of the filter model, the adaptive UKF is applied to decrease the drift estimation errors of the gyroscope by monitoring the measurements vectors and adjusting its covariance matrix on-line. The experiment results indicate that the accuracy of the calibrated MEMS gyroscopes has improved about 30%, and the filter parameters are adjusted automatically when the magnetometer measurements are deteriorated, which make the filter convergence. Furthermore, the accuracy of dynamic attitude computed by the calibrated MEMS gyroscope is smaller than 2°.