A Novel Error Correction Approach to Improve Standard Point Positioning of Integrated BDS/GPS

Area of Science:

• Satellite Navigation Systems
• Machine Learning in Geomatics
• Signal Processing

Background:

• Standard Point Positioning (SPP) using integrated BeiDou Navigation Satellite System (BDS) and Global Positioning System (GPS) faces accuracy limitations due to multiple error sources.
• Existing methods like Weighted Least Square (WLS) and Kalman filters offer partial error reduction but can be further improved.

Purpose of the Study:

• To develop and implement a novel Long Short-Term Memory (LSTM) error correction recurrent neural network approach for enhancing integrated BDS/GPS SPP accuracy.
• To reduce positioning errors originating from various sources in real-time satellite navigation.

Main Methods:

• Proposed LSTM-based algorithms: Weighted Least Square-LSTM (WLS-LSTM) and Kalman-LSTM error correction methods.
• Utilized LSTM networks to predict and correct positioning errors for subsequent epochs.
• Compared performance against traditional WLS and Kalman filter methods using measured static and dynamic data.

Main Results:

• The Kalman-LSTM method achieved a 3D point positioning error of 1.038 m, significantly outperforming WLS (3.498 m) and Kalman filter (3.406 m).
• The WLS-LSTM method also showed improvement with an error of 1.782 m.
• In dynamic scenarios, the corrected positioning error was reduced to 0.7493 m from an initial error of 3.7399 m.

Conclusions:

• LSTM-based error correction substantially enhances the standard point positioning accuracy of integrated BDS/GPS.
• The proposed Kalman-LSTM method demonstrates superior performance in mitigating positioning errors for both static and dynamic applications.
• This approach offers a promising direction for improving the reliability and precision of multi-GNSS receiver end positioning.