A navigation algorithm which combined the attention mechanism-based gated recurrent unit(GRU)neural network-assisted inertial navigation system(INS)and the global navigation sat-ellite system(GNSS)is proposed for the problem of low positioning accuracy caused by the lock-lose in the GNSS signals.The method applies a 5-point median filter to reduce random noise from the measurement signal of the inertial measurement unit(IMU),and trains a GRU neural network combined with attention mechanism to assist navigation.When the GNSS signal loses lock,the pseudo GNSS position is predicted to assist the integrated navigation,solving the problem of position divergence over time in the inertial navigation system(INS).Simulation results indicate that compared to algorithms based on multi-layer perceptron(MLP)neural networks and long short-term memory(LSTM)recurrent neural networks,the proposed algorithm exhibits lower maximum errors in northward and eastward positions in both turning and straight-line scenarios.The proposed algorithm can enhance navigation positioning accuracy when lock-lose occurs in the GNSS signals.
NETL
NSTL
万方数据
