Attitude estimation for remote sensing satellite CBERS-4 using unscented Gaussian sum filter

Abstract

The fundamental concept in a Gaussian Sum Filter (GSF) is to use a finite set of Gaussian distributions to estimate and to construct the probability density function (pdf) using Bayesian estimation approach. The goal for the GSF is approximates the predicted and posterior probabilities densities functions (pdfs) as a finite number of weighted sums of Gaussian densities distributions as has been proposed in the literature. The central idea in the Unscented Gaussian sum Filter (UGSF) is to represent non-Gaussian densities using sigma points by an Unscented Transformation. For nonlinear systems, such as attitude dynamics and attitude kinematics, the posterior pdf may not be Gaussian though, which may lead to problems in the Extend Kalman Filter (EKF) and Unscented Kalman Filter (UKF). The purpose of this research is to apply the UGSF and GSF for CBERS-4 attitude and gyros bias estimation, remote sensing satellite recently in operation. The results for attitude estimation the UGSF has a processing time 5.6 times greater than the UKF; 6.6 times less than the GSF and 19.9 times less than the Particle Filter (PF). It is possible to obtain precision in the attitude determination within the prescribed requirements using the UGSF with lower computational cost and with a smaller number of particles when compared to the standard PF.