Relying on a single method of navigation is difficult to meet the high precision requirement. Thus, the strap-down inertial navigation system (SINS) and global navigation satellite system (GNSS) integrated with complementary advantages can better meet the performance requirement. The SINS/GNSS integrated navigation system usually uses the Kalman filter to realize the data processing. When the integrated navigation system model is accurate enough, the filtering performance is good; when the navigation system model has error or change, the correction function of new measurement decrease, but the old measurement increase relatively. In order to solve the above problem, based on the SINS/GNSS integrated navigation system with the centralized Kalman filtering structure, this paper presents a new method which can process the data by estimation iterating on the gradient direction, correct the model error and improve the navigation positioning precision. The experimental results show that the new method can provide effective positioning precision for navigation systems when the navigation system model and the measurement equation have error or change, and can meet the HALE system requirements.