GPS receivers timing data processing using neural networks: optimal estimation and errors modeling.

Area of Science:

• Satellite Navigation Systems
• Signal Processing
• Artificial Intelligence

Background:

• The Global Positioning System (GPS) provides navigation and timing for military and civilian users.
• Civilian GPS applications have expanded significantly, with commercial aviation being a key area.
• High accuracy timing is crucial for certain GPS applications, but noise reduction is a challenge.

Purpose of the Study:

• To propose and evaluate five Neural Network (NN) models for reducing noise in GPS receiver timing data.
• To assess the effectiveness of these NNs in improving the accuracy of GPS timing.
• To investigate the performance of NNs using actual collected data and an experimental test setup.

Main Methods:

• Development and implementation of five distinct Neural Network (NN) architectures.
• Utilizing an experimental test setup with a Coarse Acquisition (C/A)-code single-frequency GPS receiver.
• Collecting and analyzing actual GPS timing data to evaluate NN performance.

Main Results:

• The proposed NN methods demonstrated significant noise reduction capabilities for GPS timing data.
• Experimental results showed a substantial improvement in timing accuracy.
• Root-mean-square (RMS) error in GPS timing was reduced to less than 120 nanoseconds (with SA) and 40 nanoseconds (without SA).

Conclusions:

• The evaluated Neural Networks (NNs) show strong potential for enhancing the accuracy of GPS timing data.
• These methods are effective in reducing noise, leading to high-precision timing crucial for advanced applications.
• The findings support the use of NNs for improving the reliability and accuracy of GPS timing services.