Nowadays, Autonomous Underwater Vehicles (AUV) are used in environmental studies, ocean floor mapping, and measuring water properties. Navigation of these vehicles is one of the most challenging issues due to the unavailability of global positioning system (GPS) signal underwater. Inertial navigation is a method commonly used for underwater navigation. If a low-cost Inertial Measurement Unit (IMU) is used, navigation quality will decline rapidly due to sensor inherent error. Although using a Doppler Velocity Log (DVL) speedometer sensor helps limit this error to some extent, it does not yield acceptable accuracy in low-cost IMUs. Filtering the gyro based on the AUV rotational dynamics model can improve the quality of angular velocity measurements and increase the Inertial Navigation System (INS)-DVL navigation accuracy. The presence of uncertainty in the rotational model parameters reduces the navigation algorithm’s performance. In this paper, at first, a simplified rotational model of AUV is presented, and its parameters are identified utilizing data acquired in real experiments. Then a robust Kalman filter for gyro output is proposed to enhance the navigation algorithm performance in the presence of model parameter uncertainty. To evaluate the performance of the proposed method, non-robust and robust filtered gyro outputs are used in the INS-DVL algorithm, and the results are compared with each other. Two types of the robust Kalman filter, stationary and finite horizon, are invoked. According to the field tests, navigation error decreases by 50% using the stationary robust Kalman filter compared to the non-robust Kalman filter.