Satellite attitude estimation using a two-step optimal estimator

This paper presents a new approach to recursive satellite attitude determination from vector observations that utilizes the two-step optimal estimator. It will be shown that near optimal results are obtained with very realistic noise and sensor models. The two-step estimator was introduced in 1995 as an alternative to the Extended Kalman Filter (EKF) for recursive estimation with nonlinear measurements. By breaking the measurement update process into two-steps, a linear first step using a nonlinear transformation of the desired states and a non-recursive second step minimization to recover the desired states, a near optimal estimation is possible. It has been shown that for certain systems this process recovers the global optimal solution. It will be demonstrated here that the problem of estimating the quaternion attitude representation of a satellite comes very close to this global optimum. Simulations will show dramatic improvements in performance over the traditional EKF and other attitude estimators.