Integrity, Fault Detection, and Safety-Quantifiable Localization

Accuracy is not enough for safety-critical navigation. A navigation system that gives a confident-looking answer in the wrong place can be more dangerous than one that says “I don’t know.” Safety-critical autonomy needs navigation systems that can say not only “Where am I?” but also “Is this answer safe to use?”

We develop fault detection, non-Gaussian error bounding, protection-level computation, and integrity-constrained optimization methods so navigation systems can quantify when their outputs are safe to use. Representative threads include the Principal Gaussian Overbound for Heavy-tailed Error Bounding (Yan, Zhong, Hsu, IEEE TAES 2024), Integrity-Constrained Factor Graph Optimization (Xia, Wen, Hsu, NAVIGATION 2024), and Fault Detection for Gaussian Mixture Noises in Lidar/IMU Integrated Localization (Yan et al., NAVIGATION 2025).

This pillar is the technical core of the 2023 ION Per Enge Early Achievement Award and the RGC Research Impact Fund R5009-21 (HK$4.5M, PI, 2021–2026).