An integrated navigation algorithm assisted by CNN-Informer during short-time GNSS outages

Abstract

Abstract Inertial Navigation System (INS) and Global Navigation Satellite System (GNSSs) are commonly used in integrated navigation systems. The GNSS navigation system, however, is not autonomous, and its signals are susceptible to environmental influences leading to poor navigation results. In order to improve the positioning accuracy of the INS/GNSS integrated navigation system during GNSS outages, a hybrid algorithm of Convolutional Neural Network (CNN)-Informer model-assisted INS is proposed. The CNN-Informer model combines the advantages of each of the convolutional neural network and the Informer network. It utilizes CNN to rapidly extract spatial features from input data, employs the Informer network to establish relationships between inputs and outputs, and acquires pseudo-GNSS signals to compensate for position errors in the INS. Lastly, we evaluate the proposed algorithm through six sets of experiments conducted on public datasets and real road scenarios. The comparative results demonstrate that the proposed method offers a more accurate and reliable solution during GNSS outages. Specifically, CNN-Informer can improve navigation accuracy by 81.46% over MLP and by 58.91% over LSTM during 20-second GNSS outages.