Towards Data-Driven Railway Maintenance Supported by Autonomous Inspection Robots

Licentiate thesis, 2026

Europe aims to double the use of railway transport by 2050 to support efficient and low-carbon transportation. However, increasing the pressure on an already stressed system calls for innovative and modern solutions. Integrating rail-bound autonomous robots in daily inspection, diagnosis, maintenance and repair operations would be a major leap forward. Benefits include high flexibility and efficiency in terms of cost and workforce needs. In this thesis, the concept and design of a versatile inspection, diagnostic and autonomous repair robot is presented, with a focus on onboard localization. While recent technological advances escalated the deployment of robotic solutions in numerous fields, the railway ecosystem is characterized by extensive and constraining safety regulations regarding vehicles and operations. To satisfy current requirements, special attention must be paid to achieving a robust and reliable tracking of the robot at all times, regardless of weather conditions, satellite coverage or network connectivity. This work uses custom digital maps of the French and Swedish railway networks to enable real-time path planning and navigation onboard a robot. Such maps also strongly improve the reliability of positioning approaches by constraining the problem to one-dimensional tracks. Beyond improving the quality of georeferenced data collected during inspection, a precise localization of the robot allows a safe and efficient navigation throughout the railway network. An obstacle detection system was developed to reduce collision risk. Finally, preliminary results are obtained with a prototype during field tests on a real track. These results demonstrate the feasibility of map-enabled autonomous inspection and underscore the critical role of resilient localization for real-world deployment.