AI methods for vision-based terrestrial localization

Research Project, 2026 – 2027

Modern transport systems rely heavily on satellite-based localization, making them vulnerable to disruptions such as GNSS jamming and spoofing. This project develops robust, interpretable, and infrastructure-light AI methods for localization using visual input, inertial data, and public maps such as OpenStreetMap. Focusing on diverse transport modes, i.e., pedestrians, bicycles, and trains, it builds modular algorithms for visual odometry, landmark recognition, and sensor fusion. The resulting open-source framework will support GNSS-resilient navigation, advancing safer, more autonomous, and sustainable mobility systems.