MLISB-RTK: Machine Learning Based on Inter-System Biases to Improve the Performance of RTK in Complex Environments

Area of Science:

• Geomatics Engineering
• Satellite Navigation Systems
• Machine Learning Applications

Background:

• Real-time kinematic (RTK) ambiguity resolution faces challenges with false alarms and missed detections, impacting positioning reliability.
• Existing methods for verifying RTK ambiguity fixing often rely on empirical values, limiting their effectiveness.

Purpose of the Study:

• To develop and validate a machine-learning method for correctness checking of RTK ambiguity fixing.
• To reduce false alarms and missed detections in RTK positioning.

Main Methods:

• An inter-system differential RTK model was employed, introducing the differential inter-system biases (DISB) feature.
• Machine learning classification utilized DISB, ratio value, DOP value, and residuals as features.
• The proposed method was evaluated using the SmartPNT-POS dataset.

Main Results:

• The machine-learning method significantly reduced missed detection probability by 2% and false-alarm probability by 29%.
• Positioning accuracy improved by approximately 7% compared to traditional methods.
• The DISB feature demonstrated the highest contribution rate in the classification model.

Conclusions:

• The proposed machine-learning approach effectively enhances the reliability and accuracy of RTK ambiguity resolution.
• The DISB feature is a valuable indicator for improving machine learning-based RTK verification.
• This method offers a promising solution for robust positioning in challenging environments.