An enhanced GNSS/INS navigation compensation method using LSTM-FPN for bridging GNSS outages

Abstract The integrated navigation system combining the Global Navigation Satellite System (GNSS), and Inertial Navigation System (INS) is extensively utilized for navigation and positioning, yet it encounters significant accuracy degradation in a GNSS-denied environment due to interference with the GNSS signals. To enhance the positioning accuracy of the integrated GNSS/INS during GNSS outages, this study proposes a compensation method for the GNSS/INS integrated navigation system, assisted by Long Short-Term Memory with Feature Pyramid Network (LSTM-FPN) model. This method corrects navigation errors in the INS by predicting GNSS pseudo-measurements and integrating them with INS. The LSTM-FPN enhances its adaptability to changes at different time scales in time series by extracting and fusing multi-scale temporal features, significantly improving prediction accuracy and the robustness of the model against noise and outliers. The results of real field test demonstrate that compared to pure INS, the LSTM-FPN significantly increases positioning accuracy during GNSS outages of 30 s, 90 s, and 180 s, with improvements of 90.19%, 93.03%, and 98.83%, respectively. Thus, the method effectively enhances positioning accuracy during GNSS outages.