The Application Study Of UKF And EKF On The Attitude Determination Of Satellite

Two attitude determination systems are discussed for three-axis stabilization satellite. One uses gyros and magnetometer, and the other uses gyros and star sensor. Based on the structure of"prediction and propagation", the UKF and EKF are designed for both systems, and the performances are analyzed subsequently.For UKF design, Modified Rodrigues parameter is used to describe attitude error. The main idea for time propagation is based on unscented transform.The sigma points can be computed by cholesky decomposition and propagated derectly by the nonlinear function. The mean of the sigma points is predicted mean and covariance. The attitude error then can be estimated by kalman filer using the predicted values.For EKF design, the attitude errors are represented by quaternions. The state estimation is approximated by Taylor expansion and one order truncation..The error state estimation can be computed based on the linearized model by kalman theory. Comparing the structure of two methods, the UKF apply the nonlinear function directly, so it has more advantage at accuracy and converge than the EKF.The simulation is based on three-axis stabilization satelllite in the low earth orbit. Several performances comparison between the UKF and the EKF are made through simulated examples. Simulation results indicate that the converge error and velocity of the UKF are better than the EKF,especially under the large initialization errors condition.