Handling method for GPS outages based on PSO-LSTM and fading adaptive Kalman filtering

Area of Science:

• Navigation Systems
• Artificial Intelligence
• Signal Processing

Background:

• Global Positioning System (GPS) and Inertial Navigation System (INS) integration is crucial for accurate positioning.
• GPS signal outages significantly degrade the performance of integrated navigation systems.
• Existing methods struggle to maintain accuracy during prolonged GPS unavailability.

Purpose of the Study:

• To develop a robust method for mitigating navigation performance degradation during GPS outages.
• To enhance the accuracy and reliability of GPS/INS integrated navigation systems.
• To optimize the prediction of pseudo-positions using machine learning and adaptive filtering.

Main Methods:

• A Particle Swarm Optimization (PSO) algorithm was used to optimize hyperparameters (neuron count, learning rate) for a Long Short-Term Memory (LSTM) network.
• The PSO-LSTM model predicts pseudo-positions to bridge GPS signal gaps.
• A Fading Adaptive Kalman Filter (FAKF) was employed to mitigate outliers and accumulated errors from predicted pseudo-positions, adaptively adjusting observation noise covariance.

Main Results:

• The proposed PSO-LSTM method effectively reduces positional errors during GPS outages.
• The Fading Adaptive Kalman Filter significantly mitigates the impact of observation anomalies and prediction errors.
• Compared to the Extended Kalman Filter (EKF), the FAKF improved 3D positioning accuracy by up to 23.6%.

Conclusions:

• The PSO-LSTM approach combined with FAKF offers a robust solution for maintaining high-accuracy navigation during GPS outages.
• This integrated method enhances the overall reliability and resilience of GPS/INS navigation systems.
• The findings demonstrate a significant improvement in positioning accuracy compared to conventional filtering techniques.