Nonsingular Attitude Filtering Using Modified Rodrigues Parameters

Abstract

A method to estimate the general rigid body attitude using a minimal modified Rodrigues parameters (MRP) coordinate set is presented. The singularity avoidance technique is based on the stereographic projection properties of the MRP set, and makes use of a simple mapping relationship between MRP representations. Previous work has used the MRP duality to avoid singular attitude descriptions but has ignored the associated covariance transformation. This article presents a mapping to transform the state covariance matrix between these two representations as the attitude description is mapped between the two possible MRP sets. Second-order covariance transformations suitable for divided difference filtering are also provided. The MRP filter formulation based on extended Kalman filtering and divided difference filtering is compared with a standard multiplicative quaternion Kalman filter in an example problem.