Algorithms and concepts for an onboard train localization system for safety-relevant services

This paper describes a system that enables a railway vehicle to determine its position in a track network accurately. The system does not rely on trackside hardware like balises or axle counters but is based solely on onboard sensors. It is composed out of drift-free velocity estimates of an eddy current sensor, a GNSS sensor, and a geodetic and topological track map. The paper develops an algorithm based on a probabilistic modeling that fuses the data of those sensors and determines position estimates robustly. We describe how we can treat ambiguities and stochastic uncertainty adequately and we introduce the concept of virtual balises that replace in software what is implemented by trackside balises in present train protection systems. The technique of onboard train localization is one important contribution to future train protection systems that are based on onboard sensors rather than trackside infrastructure and that are more flexible and less expensive than today's systems without loss of safety.