Satellite Attitude Determination Based on Gyro and Star Sensor
|School||Harbin Institute of Technology|
|Course||Control Science and Engineering|
|Keywords||Satellite attitude determination Extended Kalman Filter Nonlinear filter Error quaternion Semi-phsical real-time simulation|
Satellite attitude determination system is an important part of the attitude control system (ACS), and its accuracy is the key factor for the stabilization of the ACS. In general, the estimation accuracy not only lies on the performance of the hardware of the measurement devices, such as star sensor, gyro, etc, but also the attitude estimation algorithm. In this dissertation, the nonlinear filter methods are deeply investigated on theory and application for the attitude estimation from vector observations of the three-axis stabilized satellite attitude measurement system. Furthermore, to avoid the shortages existed in extended Kalman filter (EKF), several improved schemes are presented. The main contents of this thesis are as follows.Firstly, the model used for the satellite determination is set up by the accurate attitude sensor models and using quaternion. EKF algorithm is applied to modify the estimation error of satellite attitude and estimation error of gyroscope drift and information fusion. When the body rates are small, the state equation of the filter can be approximately disposed by using UEKF algorithm. The equations are uncoupled from the three-axis, as a consequence, the number of Kalman filter gains in this algorithm decreases from 18 to 6. The intention of EKF algorithm is the computation of Kalman filter gains. CGKF using a constant matrix to replace the computation of Kalman filter gains, no need to perform the calculation of state error covariance, with the obvious savings in computational effort. Finally, through analysis and comparison from the simulaton results, the above three algorithms all can gauge the performance in some conditions. Moreover, the last two algorithms can reduce large number of computation.When we can’t get the exact measurement model of the gyro, or don’t know the pre-examination knowledge about the sensors, an approach of nonlinear predictive filter based on minimum model error (MME) rule is proposed. This approach uses the quaternion which output from star-sensor to be an observation equation, forecasted and estimated the angle velocity error caused by gyro drift, then get more accurate three-axis attitude information of satellite. In addition, a modified algorithm is presented. By incorporating error quaternion into the cost function, we can estimate the system model error, and therefore the accurate attitude and angular velocities are acquired. The simulation result and the statistic of computation shows that the two algorithms all can estimate the attitude angle and compensate the error of attitude angle velocity, and compared with above, the number of computation has reduced in different degree.At last, this dissertation introduces a method of building a real-time simulation platform for satellite Guidance, Navigation and Control (GNC) system, including one host PC and two target PCs based on RT-LAB. And a super real-time simulation platform is put up to solve the bottleneck problem, which bring from software programme in the system examination. The super real-time simulation results show that the attitudue determination algorithm satisfy the real-time characteristic. This method is not only suitable for the digital real-time simulation test, but also can be extended to the semi-phsical real-time simulation test of satellite GNC system by adding some corresponding hardwares.