Model predictive filter based neural networks for INS/GPS integrated navigation during GPS outages

Aiming to improve positioning precision of the INS/GPS integrated navigation system during GPS outages, a novel neural network learning algorithm based on model predictive filter (MPFNN) for INS errors compensation is proposed. MPFNN is applied to establish a highly accurate mapping relationship when GPS works well and to predict INS errors during GPS outages. Different from traditional algorithm, model predictive filter (MPF) is established by using network weights as system state variables to optimize the network weights based on the neural network's error correction. During the training process, the model error of neural networks is corrected by compensating the deviation between the actual and target output via the MPF algorithm. Performance of the proposed method has been experimentally verified in a land vehicle navigation test. The comparison results indicate that proposed method can effectively provide high accurate corrections to the standalone INS during GPS outages.