Modified multiplicative quaternion cubature Kalman filter for attitude estimation

Abstract

Summary In this paper, a modified multiplicative quaternion cubature Kalman filter (CKF) for attitude estimation is proposed. For high‐dimensional state estimation, the CKF that uses third‐degree spherical‐radial cubature rule can provide a more accurate estimation than the unscented Kalman filter. However, for the attitude estimation in the case of larger initial conditional errors, the results may be reversed. To take full advantage of the CKF, the Lagrange cost function method is introduced to solve the quaternion weighted mean, then, the mean is used as the reference quaternion for the measurement update in the CKF framework. The choice of the reference quaternions and the quaternion update method is different from the existing literature to avoid the algorithm from failing. In addition, the unconstrained three‐component vectors represent the attitude error quaternion in the filtering algorithm, whereas the quaternion is used to perform attitude propagation. Simulation results demonstrate the better performance of the proposed modifying filter algorithm in comparison with the multiplicative extended Kalman filter, the unscented Kalman filter, and the CKF under larger initial condition errors.