Abstract: Concerning the problem of the ultra-tightly coupled global positioning system/strapdown inertial navigation system (GPS/SINS) degraded in high dynamic and weak signal environments for linear filter, a quaternion unscented Kalman filter method was proposed. Firstly, analyzing the relationship between GPS and SINS of the ultra-tightly coupled GPS/SINS navigation system, an algorithm that using the SINS to estimate and correct the carrier signal of the GPS was derived, then the state and measurement equations was built. Secondly, concerning the problems of singularity and frequent parameter transformation for Rodrigus method, based on the multiplicative quaternion error model, a weighted averaging quaternion method that using the quaternion directly was proposed. Finally, the quaternion unscented Kalman filter method was proposed for the ultra-tightly coupled GPS/SINS navigation system. The simulation results indicate that the quaternion unsecured Kalman filter method is more robust to jamming and increases the position accuracy in high dynamic.