Abstract: For the nonlinearity of the system dynamic and measurement model in Tightly Coupled Integration of LEO, this article holds research on Unscented Kalman Filter (UKF). The system model and UKF method in inertial frame are introduced. The attitude is represented by modified Rodrigues parameters instead of quaternion for its unit norm constraint, and the state is predicated by the fourth Runge-Kutta technique for satellite's high speed. The performance of UKF is compared to Extended Kalman Filter (EKF) by simulations. The results indicate that UKF is better on attitude precision than EKF by one order of magnitude, equal on position and velocity precision, but higher on computation time.