Abstract: This article presents a stable extended Kalman filter algorithm for spacecraft attitude determination by means of angular rate and vector observation. The filter using the proposed algorithm can solve the singularity in the filtering computation without any loss of system observability degree. The traditional singularity-solving method is to delete a line from the measurement equation. By comparing the system observability degree after and before the line deleting, it is proved that the traditional method tends to decrease the system observability degree. This article also proposes a method to choose the optimal deleting line to reduce the obser-vability degree loss to a minimum. The proposed projection algorithm gives a general solution to the singularity problem without observability degree loss. Theoretical analysis and numerical simulation show that an attitude filter using the projection algorithm provides more stable performance and higher estimation efficiency than one using the traditional method.

Key words: attitude estimation, celestial navigation, Kalman filtering, observability degree, system stability