Filtering Algorithm For Satellite Attitude Determination And Sensor Alignment Calibration

Satellite attitude determination system is the important part of the attitude control system (ACS), and its accuracy is the key factor for the performance of the ACS. In general, the estimation accuracy not only lies on the performance of the hardware of the measurement, but also the attitude estimation algorithm. In this dissertation, the nonlinear filter methods are deeply studied on theory and application for the attitude estimation from vector observations of the three-axis stabilized satellite attitude measurement system. Furthermore, for the shortages in extended Kalman filter, several improved schemes are presented. The main contents of this dissertation consist of the following parts:An improved unscented Kalman filter using spherical simplex unscented transformation (SUKF) is derived in an attempt to solve the attitude estimation problem with the biased gyro or not from the vector observations. As compared with EKF, no Jacobian matrix is required to calculate in UKF and better estimates of the mean and covariance of the states can be generated, which leads to better performance, but the computational costs of UKF is directly proportional to the number of sigma points which are used. While by introducing the spherical simplex UT, a better-behaved sigma point selection strategy, the number of sigma points required to the same transformation accuracy can be dramatically reduced, and accordingly, the computation burden can be avoided. In addition, with the gyro and the help of the modified Rodrigues parameters (MRPs), the proposed method can make the unit norm constraint on quaternion be satisfied. On other hand, an SUKF attitude estimator is extended to the case of the gyroless satellite with greater accuracy than attainable with an extended Kalman filter. Finally, for comparison, EKF, UKF and central difference Kalman filter (CDKF) estimators are used to gauge the performance of SUKF estimator.In order to improve the performance of EKF and UKF, the square soot unscented Kalman filter (SR-UKF) with multiplicative noise is firstly investigated. Considering the satellite nonlinear model, the SR-UKF attitude estimator is derived for gyroless satellite attitude estimation to render the state covariance with positive semi-definiteness and exceptional estimation accuracy.