Data-Driven Integrated Inertial Navigation Based on ESN Model

Global Navigation Satellite Systems (GNSS) systems fail to achieve navigation positioning in thoroughfares, viaducts, urban canyons, and tunnel environments, where the navigation signal is lost due to the particular navigation position. Inertial navigation systems, in turn, can fail to achieve satisfactory accuracy due to errors generated by the inertial measurement unit (IMU). Therefore, improving the accuracy of inertial navigation algorithms remains a challenge. They were reporting new research findings. In this paper, we develop deeply integrated inertial navigation algorithms to model and compensate for IMU errors, thereby improving the accuracy of IMU positioning. To achieve this, we have added a modeling step to the inertial navigation algorithm to learn the IMU error model during integration. First, the errors generated by the inertial navigation algorithm are trained using an Echo State Network. Second, the completed training network is depredicted. Moreover, third, the predictions are compensated into the inertial navigation results, thus integrating the inertial navigation algorithm, deep learning network, and error compensation. The results show that our deeply integrated inertial navigation algorithm can achieve much higher accuracy than inertial navigation algorithms.