Abstract: To solve the problem of irregular shape and uneven mass of small celestial body, the key technology of accurate navigation on autonomous precise landing is studied based on improved nonlinear predictive filter. First, the model of gravitational field is established in the design of navigation, and the multi-information measurement model is established based on the inertial measurement, optical camera and the velocity sensor. The nonlinear predictive filter (NPF) is improved combined with the extended Kalman filter (EKF), and the observability of the system is analyzed with the algorithm. The simulation results show that the proposed method can estimate the model error in real time under certain model error caused by gravitational uncertainty. The validity and accuracy of the improved NPF algorithm are verified comparing with the EKF algorithm.