Abstract: An improved unscented Kalman filter (UKF) algorithm based on spherical simplex unscented transformation (SSUT) was presented to satisfy the stability, precision and real time requirements of spacecraft autonomous navigation. The S SUT reduces the sigma points for the unscented transformation. Compared with the traditional UKF, the computational efficiency of spherical simplex unscented Ka lman filter (SSUKF) was improved, without sacrificing the estimation precision. A ccording to the similar computation process of UKF and extended Kalman filter (E KF), the combined Kalman filter based on SSUKF and EKF was designed. This algori thm can switch adaptively between the SSUKF and EKF through a mode switching fun ction which is the criterion of estimation errors. It solves both the inefficien cy problem of UKF and the instability problem of EKF. Simulation results show th at the combined Kalman filter integrates the advantages of SSUKF and EKF, improv es the computational efficiency, insures the estimation precision and robustness , therefore it is more suitable for the spacecraft autonomous navigation system.