LSTM assisted in vehicle GNSS/INS integrated navigation algorithm and performance analysis

Aiming at the problem that the positioning accuracy of the vehicle mounted GNSS/INS integrated navigation system declines or even diverges when the GNSS signal is unlocked, a new algorithm based on long short memory (LSTM) neural network assisted integrated navigation is proposed to improve the positioning accuracy and achieve reliable, continuous and stable positioning. The experiment was conducted on mobile integration platform, and the results showed that when the GNSS signal lost lock for 30 seconds, the maximum position error of the LSTM assisted integrated navigation system in the east and north directions decreased by 77.45% and 17.39%, respectively, and the root mean square error (RMSE) decreased by 79.53% and 42.36%, respectively; When the GNSS signal loses lock for 100 seconds, the maximum position error values of LSTM assisted GNSS/INS in the east, north, and sky directions decreased by 60.07%, 98.30%, and 84.65%, respectively, while RMSE decreased by 61.96%, 97.98%, and 84.65%. LSTM assistance greatly improves the navigation performance of the onboard GNSS/INS integrated navigation system.