Online update of GRU network-assisted GPS/INS positioning compensation algorithm

Traditional neural network-assisted combinatorial navigation learning models use a fixed training dataset, which results in the model not being able to capture the features and changes of the latest navigation data, the cumulative error increases in the case of longer interruptions of the satellite signal, and the prediction performance of the model deteriorates. To overcome this problem, a positioning compensation algorithm based on online updating gated recurrent unit (GRU) neural network is proposed, which dynamically updates the training set in a sliding graph window, trains the model online and fine-tunes the parameters to achieve real-time prediction of position increments. The experimental test results show that the proposed method can effectively improve the positioning accuracy, still have more reliable navigation performance in the case of longer interruptions, and have good generalization ability.