Aimed at the problem that the satellite signals in integrated navigation have an effect on error, a new robust Kalman filter algorithm based on chi-square on test and Newton interpolation is proposed. The algorithm is to improve the positioning accuracy of chi-square test to the fault measurement value by taking the residual error of the measurement value disobeyed the Gaussian distribution with the mean value being zero as a theoretical basis and the simulation experiment as a basis through designing a method to select the threshold of chi-square test to meet the significance level requirement. On the basis mentioned above, the Newtonian interpolation polynomial is constructed by using the observed values at the past time and donated to get the new measured values at the current time, and measurement noise is improved by introducing improvement factor, making measurement noise more close to normal value, and finally improving navigation accuracy. The experimental results show that the algorithm proposed in this paper reduces the position error by about 75％ and the speed error by about 80%, compared with the general filtering algorithm, and can also effectively suppress the influence of observed faults and improve the reliability of integrated navigation system.