Research on GNSS/DR method based on B-spline and optimized BP neural network

To improve the positioning accuracy of Micro Electromechanical System (MEMS), Inertial Measurement Unit (IMU), and Dead Reckoning (DR) navigation under a short-term loss of Global Navigation Satellite System (GNSS) signal, we proposed a neural network algorithm based on B-spline time synchronization. Our method improves the performance of MEMS inertial navigation positioning accuracy. In this paper, the GNSS discrete value is converted into a continuous value, the angular velocity and acceleration at any time are obtained by derivation. An algorithm based on the B-spline point acceleration and angular velocity criterion is designed to calculate the time delay between GNSS and MEMS IMU. In this paper, B-spline is used to convert the position information of GNSS into angular velocity and acceleration information. This greatly simplifies the model, while also suppressing inertial navigation errors. Finally, a method of varying learning rates is proposed to reduce training time. We compared other methods, the vehicle experiments show that the average positioning accuracy of pure IMU algorithm is 140.3m (RMSE) when GNSS is lost for one minute and the average travel distance is 729m, the average positioning accuracy of Extended Kalman Filter (EKF) is 65.34m, the average positioning accuracy of MLP algorithm is 58m, The average positioning accuracy of our method is 3.02m. High precision positioning in a short time is realized and the effectiveness of the algorithm is verified, the accuracy of our method is the state-of-the-art.