1. School of Astronautics, Beijing University of Aeronautics and Astronautics�� Beijing 100083, China;
2. School of Physical Geography and Information Technology, China University of Geoscience�� Beijing 100083, China
A multi-model Kalman filter based on GA(genetic algorism) was designed. When the parameter is imprecise, estimation results of Kalman filter are false.Method used in this paper can estimate the best parameters on line,so as to get true estimation results.At the end of this paper the application to integrated navigation system was studied with simulation. The results of simulation show that as compared with single-model algorithm, this method not only can improve accuracy and reliability of integrated navigation system greatly,but also can identify the noise statistical models of inertial elements in the integrated navigation system relatively well.