A navigation method based on neural process and Kalman filtering and related device

By combining neural processes and Kalman filtering navigation methods and utilizing gyroscope and accelerometer data, the problem of insufficient navigation parameter estimation accuracy after GNSS lock-up is solved, achieving high-precision navigation parameter prediction and real-time tracking of attitude changes.

CN116026325BActive Publication Date: 2026-06-16XI AN JIAOTONG UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
XI AN JIAOTONG UNIV
Filing Date
2023-02-16
Publication Date
2026-06-16

AI Technical Summary

Technical Problem

Existing neural network-based navigation methods for moving vehicles struggle to achieve high-precision long-term navigation parameter estimation after GNSS lock-on loss, especially with decreased prediction accuracy during attitude changes, and are susceptible to measurement noise and model uncertainties.

Method used

By combining neural processes and Kalman filtering, and using gyroscope and accelerometer data, navigation parameters are jointly estimated through a neural process model and a Kalman filter. LSTM is used to process time-series data to establish a navigation parameter prediction model, and attitude information is updated through Kalman filtering to achieve high-precision navigation parameter estimation.

Benefits of technology

In the case of GNSS lockout, long-term, high-precision estimation of multi-dimensional coupled navigation parameters of moving vehicles is achieved, which improves the anti-interference capability and prediction accuracy of the navigation system and enables real-time navigation that adapts to attitude changes.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN116026325B_ABST
    Figure CN116026325B_ABST
Patent Text Reader

Abstract

The application discloses a navigation method based on neural process and Kalman filtering and a related device, and comprises the following steps: solving navigation data of a moving carrier velocity and attitude, and combining the navigation data with gyroscope and accelerometer data to form a training data set; grouping data in the training data set, and establishing an NP model of a navigation method; selecting the training data to train the NP model until a training result converges, and obtaining a navigation parameter prediction model; updating attitude information of the moving carrier, inputting the updated attitude information into the navigation parameter prediction model, and estimating navigation parameters at a next moment; and repeatedly updating the attitude information of the moving carrier until a navigation task is completed or GNSS is restarted. The application combines LSTM and NP, and reduces the data format dependency of NP. The application combines NP and KF, and inputs the navigation parameters corrected by KF into NP for prediction in a next period.
Need to check novelty before this filing date? Find Prior Art