Research On Nonlinear Filtering Of Satellite Attitude Determination

EKF is a kind of familiar nonlinear filtering algorithm which is applied widely in the satellite attitude determination. However, there is some deficiency especially when it is used for serious nonlinear attitude system, which will degrade the estimation precision. So some kind of typical nonlinear filtering methods for example UKF, PF are proposed one after the other and begin to be applied to estimate satellite attitude. In this dissertation, a few nonlinear filtering methods of attitude determination are deeply researched and the modified methods are represented. The primary work in this dissertation is as follows.1. The attitude representation and mathematic model are taken as the main clue, some important basic theories and key technologies are introduced detailedly. Subsequently after introducing the nature of satellite attitude determination problem and attitude sensor measurement model, the system model equations based on quaternion and the Modified Rodrigues parameters are described, which provides a theoretical reference for further research.2. EKF is a kind of efficient method of solving nonlinear problems, however the linearization error is introduced in the process of linearization by Taylor series expansion. So the nonlinear predictive filter is used to estimate the model error which is generated in the linearization, with that the statistical characteristics of model error is taken into account in the standard EKF to make the linearization model more accordant with the actual situation.3. UKF is advised to estimate satellite attitude based on Modified Rodrigues parameters. When the observations contain gross errors, an new robust UKF algorithm grounding on statistical properties of residuals is proposed, which is compared with the robust UKF grounding on equivalent covariance matrix through simulation examples. The results show that two methods can effectively resist the observation outliers, but the new algorithm can improve the adaptive ability of the system better.4. IUKF is advised to estimate satellite attitude based on Modified Rodrigues parameters that follows the introduction of principle of IUKF. At the same time, after analyzing the IUKF algorithm, an improved IUKF with stronger robustness is proposed, which is validated through a experiment.5. This paper studies satellite attitude estimation based on quaternion using UKF and proposed UKF attitude estimation algorithm that can deal with colored observation noise, in which polynomial division is proposed to transform colored noise to virtual white noise,6. Following the description of particle filter and its main issues, particle filter and unscented particle filter is proposed to estimate satellite attitude based on Modified Rodrigues parameters. When the attitude priori information can hardly be achieved or the initial attitude error is large, the particles can be sampled according to the characteristics of the Euler angles. And then, PF and UPF is used to estimate satellite attitude based on Modified Rodrigues parameters that is achieved by the convert between Euler angles and MRPs.