It’s necessary to consider the impact of DVL measurement outliers and its time-varying noise characteristic on navigation accuracy in SINS/DVL integrated navigation. To address this problem, a sliding window robust adaptive Kalman filter for SINS/DVL integrated navigation is proposed in this paper. Firstly, the system error model is established based on psi angle state transformation. The force item of its velocity is replaced by constant gravity term, so that the accuracy of the model is improved by reducing the speed error caused by the force item. Then the innovation sequence is used to verified outliers. If the outliers exist, the sliding window data is used to correct the innovation sequence and adaptive Kalman filter algorithm with attenuating factor is utilized for estimate observation noise. The experimental results show that the new error model and sliding window robust adaptive Kalman filter can effectively suppression the position error divergence. The position error can be controlled in 1.05%D in 2h’s SINS/DVL integrated navigation. The navigation position accuracy is enhanced by 39.66% compared to Kalman filter.