Error restraining method for SINS based on Bagging model

To improve the positioning accuracy of the integrated SINS/GPS navigation system in land vehicle during GPS outages, an ensemble learning algorithm (Bagging model) is proposed. Compared with traditional single neural network, the proposed algorithm can deeply study internal relations among SINS error and further improve the navigation performance. Its error restraining scheme based on intelligence algorithm is proposed, which trains the model by using integrated navigation data when GPS works well, and predicts the navigation system position increment during GPS outages. Experiments are made to test the performance of the proposed algorithm by using GPS and SINS data collected in navigation tests of land vehicle, which show that the Bagging method can restrain the SINS error during GPS outages, and the position accuracy is improved by 49% in 5min and by 41% in 15min, compared with those by BP model.