Gaussian sum particle filter for spacecraft attitude estimation

Abstract

A novel and efficient spacecraft attitude estimation method using Gaussian Sum Particle Filter is proposed based on vector observations. The attitude is represented by quaternion, and the local error is represented by the modified Rodrigues parameters. Gaussian mixture model is used to approximate the posterior density of the state, and the state update is carried out by sampling based methods. Meanwhile Expectation Maximization algorithm is introduced to avoid the collapsing of Gaussian mixture terms. The efficiency of Gaussian sum particle filter estimator for spacecraft attitude is validated by numerical simulation. The simulation results show that Gaussian mixture particle filter is superior to Particle Filter and Unscented Kalman Filter for spacecraft attitude estimation.