Abstract: There always are unknown inputs in the dynamic model of high-speed Mars vehicle during the Mars entry phase because of the atmospheric density, aerodynamic parameters, gust and sand storm, etc. Due to the effect of the unknown inputs, the traditional filtering methods may produce greater errors. This paper estimates the states of the Mars entry vehicle by using a new self-calibration extended Kalman filter, and successfully eliminates the effect of the unknown inputs. Numerical simulation shows that this self-calibration method may effectively improve navigation accuracy.