A Train Factor Graph Fusion Localization Method Assisted by GRU-IBiLSTM for Low-Cost SINS/GNSS

Area of Science:

• Navigation Systems Engineering
• Artificial Intelligence in Transportation
• Geomatics Engineering

Background:

• Strapdown Inertial Navigation System (SINS)/Global Navigation Satellite System (GNSS) integration is crucial for railway positioning.
• Conventional filtering methods struggle with historical data and fail during GNSS outages.
• Existing solutions for GNSS outages often lead to significant error accumulation.

Purpose of the Study:

• To develop an enhanced data utilization framework for railway positioning systems.
• To improve navigation accuracy and continuity during Global Navigation Satellite System (GNSS) signal outages.
• To overcome the limitations of traditional filtering approaches in dynamic environments.

Main Methods:

• Implementation of a factor graph optimization (FGO) framework for improved data efficiency.
• Integration of a Gated Recurrent Unit (GRU) and Improved Bidirectional Long Short-Term Memory (IBiLSTM) network.
• Generation of pseudo-GNSS observations using a hybrid neural network to bridge GNSS outages.
• Evaluation using both simulation and onboard vehicle data under GNSS-denied conditions.

Main Results:

• The proposed GRU-IBiLSTM network significantly reduced horizontal root mean square error (RMSE) by 49.22% (simulation) and 36.24% (onboard vehicle) compared to conventional methods during outages.
• Subsequent factor graph optimization further improved accuracy, reducing RMSE by an additional 46.67% (simulation) and 35.31% (onboard vehicle).
• The methodology demonstrated robust performance in maintaining positioning accuracy and navigation continuity.

Conclusions:

• The developed factor graph optimization framework with a hybrid neural network offers a robust solution for train positioning.
• This approach effectively mitigates positioning errors and ensures reliable navigation during Global Navigation Satellite System (GNSS) outages.
• The findings highlight the potential for advanced AI techniques to enhance the resilience of navigation systems in challenging environments.