A Decoupled-DCM Orientation Estimator for Eliminating Influence of Magnetic Interference on Roll Angle and Pitch Angle

For mobile robots, attitude information plays a crucial role in mobile robot applications. However, the real-time positioning of robots is often hindered by issues such as robot motion, local magnetic field interference, and coupled attitude estimation. This paper proposes a double-layer attitude estimation algorithm based on error states, utilizing decoupled Direction Cosine Matrix (DCM) as the output. The algorithm addresses the limitations of traditional attitude calculation methods, which suffer from large nonlinear errors and susceptibility to magnetic interference. By employing an exponential mapping in the Lie algebra, the decoupled attitude errors are updated to the previous moment state. Moreover, a novel attitude angle decoupling method of DCM is introduced, enabling a new representation of pitch angle, roll angle, and azimuth angle. This representation retains the advantages of the direction cosine matrix while reducing the dimensionality of attitude angle representation and the nonlinearity of the observation equation. The paper further presents the "multiplicative update" and "covariance correction step" derived from the newly proposed decoupled attitude angle representation within the manifold space filtering. These concepts are applied to both Extended Kalman Filter (EKF) and Unscented Kalman Filter (UKF) frameworks. Static and dynamic tests demonstrate that both filtering algorithms effectively eliminate the influence of magnetic interference on pitch and roll angles, while achieving comparable accuracy in attitude angle estimation to state-of-the-art algorithms.