Abstract: When GPS is in good observation condition, how to improve the reliability, the continuity and the precision of the integrated navigation is the goal of many researchers. In tight integarted navigation based on pseudorange and doppler observation, the precision is poor and the modification of the attitude errors is not accurate for poor observation precision of pseudorange. So a two-step adaptive robust Kalman filtering based on the observability of the parameters is presented. First the process of the tight integration is given and then the formulas and the approches of the new method are deduced and analyzed. Finally an actual calculation is given. It is shown that campared with tight integration, the two-step adaptive robust Kalman filtering can control the disturbances of the state and the outliers of the observation. And the navigation precision does not decrease while the integration period becomes longer and the INS errors become bigger. The INS errors can be rightly estimated and the precision of attitude angles is improved.