Real-time predictions for Light rail train systems

Public transport operations are subject to inherent uncertainty. Instantaneous vehicle positioning data facilitates the development of prediction schemes that could improve the accuracy and reliability of real-time information (RTI). There is lack of research on RTI for Light rail train (LRT) systems. LRT are characterized by driving regimes that depend on rolling stock and infrastructure specifications. This paper develops and tests two prediction schemes for rail-bound systems which are based on constructing link-specific speed profiles. The prediction schemes are applied for a LRT line in Bergen, Norway. The performance of the currently deployed scheme and alternative schemes is compared using 6 months empirical vehicle positioning data. The results indicate that the proposed methods improve the reliability of the prediction scheme and increase the share of errors smaller than 1 minute to 89%, up to par with metro systems.