Square-root quaternion cubature Kalman filtering for spacecraft attitude estimation

Abstract A novel quaternion estimator called square-root quaternion cubature Kalman filter is proposed for spacecraft attitude estimation. The filter approach uses a gyro-based model for quaternion propagation and a reduced quaternion measurement model to substantially reduce the computational costs. The process and measurement noises of the system model exhibit the same kind of linear state-dependence. The properties of the state-dependent noises are extended and more general expressions for the covariance matrices of such state-dependent noises are developed. The new filter estimates the quaternion directly in vector space and uses a two-step projection method to maintain the quaternion normalization constraint along the estimation process. The square-root forms enjoy a consistently improved numerical stability because all the resulting covariance matrices are guaranteed to stay positive semi-definite. Extensive Monte-Carlo simulations for several typical scenarios are performed, and simulation results indicate that the proposed filter provides lower attitude estimation errors with faster convergence rate than a multiplicative extended Kalman filter, a quaternion Kalman filter, and a generalized Rodrigues parameters (GRPs)-based cubature Kalman filter for large initialization errors.