Improving Real-Time Position Estimation Using Correlated Noise Models.

Area of Science:

• Geomatics Engineering
• Signal Processing
• Navigation Systems

Background:

• Accurate real-time Global Positioning System (GPS) location inference and uncertainty quantification are critical for navigation.
• Existing Kalman Filter (KF) approaches often assume Gaussian noise, which can lead to underestimation of position uncertainties.

Purpose of the Study:

• To develop and evaluate novel algorithms for real-time GPS location estimation and uncertainty quantification.
• To compare the performance of different algorithms, including those with enhanced noise models.

Main Methods:

• Algorithms were tested using GPS data from the Southern Hemisphere at various latitudes.
• The Ornstein-Uhlenbeck (OU) noise model was integrated into an enhanced Kalman Filter (KF).
• Performance was ranked using the log-score rule; dilution-of-precision parameters were also evaluated.

Main Results:

• The enhanced KF with the OU noise model demonstrated superior performance in capturing autocorrelated process noise.
• This approach significantly outperformed a standard KF with a Gaussian noise model.
• GPS dilution-of-precision parameters provided minimal benefit for uncertainty quantification.

Conclusions:

• The enhanced KF with an OU noise model offers accurate real-time GPS positioning and uncertainty quantification.
• The method is computationally suitable for embedded systems and sensor fusion applications.
• It provides a robust foundation for integrating data from complementary sensors like accelerometers and magnetometers.