Autonomous alignment of continuous railway track geometry recordings from in-service vehicles

In-service railway vehicles are increasingly used for autonomously monitoring railway track condition with no disruption to passenger and freight timetables. However, comparison of track geometry recordings with inspection measurements from Track Recording Vehicles (TRVs) can be challenging due to longitudinal misalignment of recording data. Following review of existing literature on challenges of recording alignment, a method for autonomously correcting the longitudinal alignment of track geometry monitoring recordings with reference inspection data is presented which makes use of Hidden Markov Models (HMMs). This methodology is robust to estimation errors and can produce global alignment to known track sections with minimal constraints provided. Recordings taken from a section of high-speed rail line in the United Kingdom validate the methodology. Results indicate that one journey per 32 hours can be correctly and autonomously aligned to the reference, enabling more frequent monitoring of track geometry degradation at a lower cost than monthly TRV inspections can permit. The technique scales to larger fleets of instrumented vehicles to provide wider and more frequent network coverage.