An algorithm for precise localization of measurements in rolling stock-based diagnostic systems

Abstract

The paper describes the development and the validation of an algorithm for the geo-localization of trains, specifically designed for diagnostic applications of rolling stock and infrastructure. The algorithm exploits two signals commonly available on commercial trains, namely odometry and GPS, combined to overcome the drawbacks they both show when adopted individually. The algorithm’s flow consists of a map matching procedure for the projection of GPS acquisitions on digital maps of the railway line, followed by a robust fit to correlate the map-matched data to the odometry data. The system is specifically conceived to be adopted in rolling-stock based diagnostic system, to correlate the diagnostic indices to the actual position where they are gathered along the line. It is specifically designed to operate on double-track railways, even if its adoption in multi-track scenario typical of commuter rail would be possible by means of further integration with balise signals or other train control data. Results from track tests on a high-speed train with instrumented pantographs and bogies show that the developed algorithm allows to obtain a good repeatability of diagnostic indicators collected during repeated runs of the train along the same track. This result set the basis for the automation of data analysis in a wayside server, and for the computerization of the comparison of data belonging to homogeneous track sections, so as to perform a trend analysis.