Abstract: For the vehicle mounted global navigation satellite system (GNSS)/strapdown inertial navigation system (SINS) integrated navigation system, aiming at the problem of gradual divergence of longitudinal position error of SINS assisted by velocity constraint when GNSS fails and SINS works alone, a vehicle SINS positioning algorithm assisted by velocity constraint based on neural network madification is proposed. The radial basis function (RBF) neural network is used to predict the correction coefficient of SINS longitudinal position error, so as to improve the positioning accuracy of SINS when working alone. In addition, an adaptive filtering algorithm for real time measurement noise estimation with limited memory index weighting is proposed. The vehicle tests are carried out under artificially setting GNSS failures and real tunnel scenarios. The results show that the proposed algorithm can correct the longitudinal position error of SINS online without stopping. Compared with the conventional algorithm combining velocity constraint and Kalman filter, the positioning accuracy of vehicle SINS under GNSS failure is effectively improved.

Key words: Strapdown inertial navigation system (SINS), Velocity constraint, Neural network, Adaptive filter