GAN-FDSR: GAN-Based Fault Detection and System Reconfiguration Method

Area of Science:

• Navigation Systems Engineering
• Artificial Intelligence in Navigation
• Signal Processing for GNSS/INS

Background:

• Integrated Global Navigation Satellite System (GNSS)/Inertial Navigation System (INS) are critical for reliable positioning.
• Ensuring the integrity of GNSS/INS data through fault detection and exclusion is paramount.
• Existing fault detection methods may struggle with subtle or gradual system faults.

Purpose of the Study:

• To propose a novel fault detection and system reconfiguration scheme (GAN-FDSR) for tightly coupled GNSS/INS.
• To enhance the detection sensitivity for small-amplitude and gradual faults.
• To improve the overall positioning accuracy and reliability of integrated navigation systems.

Main Methods:

• Reconstruction of raw pseudo-range data in phase space to capture chaotic characteristics and non-linearity.
• Utilizing Generative Adversarial Networks (GANs) to calculate generation and discrimination scores for fault detection.
• Dynamic system reconfiguration based on the relative differential precision of positioning (RDPOP) of faulty satellites.

Main Results:

• The GAN-FDSR method significantly improves detection sensitivity for subtle and gradual faults.
• The proposed scheme effectively reduces positioning errors during fault conditions.
• Simulation experiments validate the enhanced fault detection performance and positioning accuracy.

Conclusions:

• The GAN-FDSR scheme offers a robust solution for fault detection and exclusion in tightly coupled GNSS/INS.
• Phase space reconstruction enhances the model's ability to learn non-linear data characteristics.
• Dynamic reconfiguration based on RDPOP ensures optimal system performance under fault scenarios.