Dynamic Modeling and Its Impact on Estimation Accuracy for GPS Navigation Filters.

Area of Science:

• Navigation Systems
• Estimation Theory
• Signal Processing

Background:

• Kalman filter (KF) and extended Kalman filter (EKF) algorithms rely on process models that often diverge from real-world conditions, degrading real-time navigation performance.
• Inaccuracies in dynamic modeling are a primary cause of divergence issues in Global Positioning System (GPS) navigation.

Purpose of the Study:

• To address divergence issues in GPS navigation systems caused by dynamic modeling inaccuracies.
• To introduce and evaluate fictitious process noise as a solution to mitigate EKF divergence.
• To enhance the observability and controllability of GPS navigation filters.

Main Methods:

• The study employs the extended Kalman filter (EKF) with the introduction of fictitious process noise.
• Covariance estimation and tuning factors are utilized to improve filter performance.
• Simulations include various motion scenarios: stationary receivers, straight-line trajectories (constant/varying speeds), and turning trajectories.
• Position (P), position-velocity (PV), and position-velocity-acceleration (PVA) models are evaluated using simulated receivers.

Main Results:

• The inclusion of process noise enables the EKF to adapt to changes in speed and direction without explicit dynamic modeling.
• Simulations demonstrate improved vehicle trajectory tracking compared to Ordinary Least-Squares (OLS) methods.
• The EKF with fictitious process noise shows enhanced statistical reliability and real-world performance across diverse scenarios.

Conclusions:

• Fictitious process noise is an effective remedy for mitigating divergence in EKF-based GPS navigation systems.
• The proposed approach offers a systematic method to improve estimation accuracy and broaden the applicability of navigation systems.
• This research provides valuable insights for enhancing the robustness and precision of GPS navigation filter designs.