Particle Filtering and Its Application in Satellite Orbit Determination

Focusing on nonlinear/non-Gaussian filters and based on analysis of the sampling-importance-resampling algorithm, this paper examines principles and application of particle filter (PF) in satellite orbit determination. There may be large error of initial estimation and non-Gaussian distribution of state and measurement in satellite orbit determination. To solve the two problems, an improved PF algorithm is put forward in this paper. Depending on data features, a multinomial re-sampling strategy is used for filtering and self-adaptively determines the optimal number of particles. Performance of PF under different parameters is also illustrated. Furthermore, performance of PF is compared with that of EKF and UKF in simulation experiments of satellite orbit determination. Theoretical analysis and simulation demonstrate that PF is more advantageous than EKF and UKF in operation and parameters adjusting PF in accordance with data features can promote filter speed and precision.