Variational Iterative Filter for Orbit Estimation
Xiaoxu Wang, Zhengya Ma · 2019
A nonlinear filter termed variational iterative filter (VIF) is presented in the paper. The measurement likelihood is modeled subtly as a Gaussian regression process by introducing the compensation parameters. Afterwards, assume the conjugate priori information of compensation parameters with Gaussian-Wishart distribution. Such assumption aims to make the posterior estimates conjugated with the prior for facilitating the computation. Then in variational Bayesian (VB) framework, the state estimation and compensation parameters identification are found alternately and iteratively by minimizing the Kullback-Leibler (KL) divergence. Finally, to evaluate the effectiveness of VIF, it is compared with Kalman filters by estimating a low-Earth spacecraft's state with the ground radar.