An Approach for Improving Pure-Inertial Navigation Accuracy by Optimizing the Pre-Integration Window of Deep Learning Model

The application of underwater inertial navigation in autonomous underwater vehicle ( AUV ) is very important, but the cumulative error of inertial navigation in complex sea conditions is large, which seriously affects its positioning accuracy. In order to solve this problem, this paper proposes a deep learning pure inertial navigation method based on preprocessing time window optimization, and focuses on the influence of the length of preprocessing time window on the accuracy of motion calculation. The purpose is to improve the adaptability and robustness of the inertial navigation system in different motion states by optimizing the length of the preprocessing time window. In order to verify the effectiveness of the proposed method, the researchers conducted sea trials, during which navigation data under different motion modes were collected. The experimental results show that the proposed method can effectively reduce the cumulative error of inertial navigation under complex sea conditions compared with the original method, which verifies the effectiveness and universality of the method, and has the value of popularization and engineering application.