Iterated Multiplicative Extended Kalman Filter for Attitude Estimation Using Vector Observations

This paper proposes an iterated multiplicative extended Kalman filter (IMEKF) for attitude estimation using vector observations. In each iteration, the vector-measurement model is relinearized based on a new reference quaternion refined by the attitude-error estimate. An implicit reset operation on the attitude error is performed in each iteration to obtain the refined quaternion. With only a little additional computation burden, the IMEKF can much improve on the performance of the MEKF. For large initialization errors, the IMEKF performs even better than the unscented quaternion estimator but with much smaller computational burden. Numerical results are reported to validate its effectiveness and prospect in spacecraft attitude-estimation applications.