Abstract: GNSS／INS navigation positioning results for dynamic environment are often affected by the gross error problem, a kind of GPS／BDS double system RTK／INS tightly-couplede integration algorithm based on robust kalman filter is proposed. According to variance expansion model, robust Kalman algorithm is established to obtain GNSS/INS tightly integrated robust solutions, and through two different regions. The experimental results show that the navigation accuracy of this method is improved by 1.4～4.6 cm, 0.7～9 cm, 1.5～2 cm, and the success rate of ambiguity fixing is increased by 10.3～25.6% compared with the traditional methods in three directions of N, E and D, respectively. The navigation accuracy and reliability of this method are improved significantly, and the vehicle navigation accuracy and reliability in dynamic environment are improved. The application of automatic driving has certain theoretical reference and practical value.