A method based on CNN-BiLSTM for UAV navigation error compensation in GNSS denied environment

When the Global Navigation Satellite System (GNSS) signal is unavailable, the performance of GNSS/Inertial Navigation System (INS) integrated navigation system will be seriously degraded. In order to compensate the performance of GNSS/INS integrated navigation system, an error compensation method based on Convolutional Neural Network (CNN) combined with Bidirectional Long Short-Term Memory (BiLSTM) network is proposed. In the proposed method, the Inertial Measurement Unit (IMU) dynamics measurement of Unmanned Aerial Vehicle (UAV) is used as the input, the GNSS high-precision navigation information is used as the label. CNN-BiLSTM establishes the relationship between IMU dynamics measurement and GNSS high-precision navigation information through the training of a large number of input and label data, since it can achieve more accurate prediction of position increment and velocity increment. At last, the experimental results show that when the GNSS signal is unavailable, CNN-BiLSTM obtains the lower average RMSE and MAE of prediction error and the higher prediction fitting degree compared with CNN-LSTM and LSTM.