Global Unscented Attitude Estimation via the Matrix Fisher Distributions on SO(3)

TL;DR

This paper introduces a global probabilistic attitude estimation method using matrix Fisher distributions on SO(3), employing an unscented transform for improved accuracy during complex maneuvers.

Contribution

It develops an intrinsic, global Bayesian attitude estimation framework that effectively handles large initial errors and uncertainties without local coordinates.

Findings

Successfully manages large initial errors and uncertainties.

Provides accurate attitude estimates during complex maneuvers.

Introduces an unscented transform for matrix Fisher distributions.

Abstract

This paper is focused on probabilistic estimation for the attitude dynamics of a rigid body on the special orthogonal group. We select the matrix Fisher distribution to represent the uncertainties of attitude estimates and measurements in a global fashion without need for local coordinates. Several properties of the matrix Fisher distribution on the special orthogonal group are presented, and an unscented transform is proposed to approximate a matrix Fisher distribution by selected sigma points. Based on these, an intrinsic, global framework for Bayesian attitude estimation is developed. It is shown that the proposed approach can successfully deal with large initial estimator errors and large uncertainties over complex maneuvers to obtain accurate estimates of the attitude.