Research on Three-Axis Stabilized Satellite Attitude Determination Based on Particle Filter
|School||Harbin Institute of Technology|
|Course||General and Fundamental Mechanics|
|Keywords||Attitude Determination Star sensor Extended Kalman filtering Particle filter|
Satellite attitude information is an important factor to affect the performance of the whole attitude control system. Its accuracy depends not only on the performance and accuracy of the attitude measurement system hardware configuration, determination algorithm is closely related with the gesture. The star sensor is a vector attitude sensor, aerospace engineering in highest precision attitude sensors. In this paper, the star sensor and gyro targeted as a basic configuration consisting of three-axis stabilized satellite attitude determination system, made an intensive study of nonlinear filtering techniques for satellite attitude determination. Mainly to complete the work of the following aspects: First, a more comprehensive and systematic summary of the various postures description, including their definitions, calculation rules, kinematic equations, translation relations and the advantages and disadvantages. Complete satellite attitude motion model based on quaternion Modified Rodrigues parameters. Secondly, the use of quaternions as the description of the satellite attitude measurement model for focal plane model, attitude determination algorithm is derived based on the combination of the star sensor and gyro extended Kalman filter. Numerical simulations show that the extended Kalman filter algorithm has better performance in a smaller initial error, can not be guaranteed to converge when the initial error is larger. In order to solve the larger initial error and non-Gaussian distribution, the focus on the application of the particle filter in satellite attitude and attitude rate determination. Have gyro and gyro particle filter SIR algorithm for posture and attitude angular velocity determined. This algorithm is based on Monte Carlo simulation with random particles to approximate the probability distribution of the state vector. Initial posture distribution is uniform, Rodrigues parameters as satellite attitude described. To illustrate the impact of the resampling particle filter accuracy, a comparison of three different resampling strategy. Particle filter algorithm like poverty issues, using the the particle coarsening method to increase the diversity of the particles. Focal plane and starlight vector two measurement models were used in the simulation. Simulation results show that the particle filter SIR algorithm in the case of using the number of particles of 2000, although in a small deviation did not exhibit superior performance, but in a large initial error and non-Gaussian distribution of the cases have good convergence properties. Moreover, two different attitude measurement model particle filter also has a very good performance. In the same time, can also be found using the different resampling strategy has little effect on the final result.