INS/GNSS integrated navigation algorithm based on CNN-LSTM-Attention

Abstract GNSS signals are very fragile. When the receiver is located in a tunnel or under a tall building in a city, GNSS signals are easily blocked, resulting in GNSS loss of lock. Therefore, an integrated navigation algorithm based on CNN-LSTM-Attention is proposed. First, when the GNSS signal is intact, important time series features are captured using CNN and attention mechanism, which improves the performance of traditional LSTM networks in processing strongly coupled navigation data. Then the CNN-LSTM-Attention prediction model is trained with the combined navigation data, and the inertial navigation error model is obtained. When the GNSS signal is interrupted, the trained prediction model is used to predict the inertial navigation error information, thereby suppressing the inertial navigation error. Finally, simulation comparison experiments are carried out. The conclusion shows that this method can solve the combined navigation problem when GNSS loses lock.