Global Sensitivity Analysis for the evaluation of the effects of uncertainty of transport demand and passenger behavior on planning railway services with variable train composition

In the last decades, the rise of automation and communication technologies in railway sector has boosted more and more opportunities to design new and innovative solutions to improve the trip experience of people. Among them there is the so-called virtual coupling, a promising framework where trains, automated and connected, can split and merge dynamically without the need of human effort and a mechanical connection, paving the way for new planning approaches enabled by the introduction of a variable train composition. The aim of this paper is to propose a methodology, based on a Global Sensitivity Analysis, for evaluating the effects of the uncertainty of the input data on a model aimed at optimally planning urban rail services with variable train composition. In particular, the sensitivity to transport demand and passenger behavior in boarding trains has been analyzed. Results allow to determine the input parameters that, being more relevant, should be estimated with higher attention in a real-world application and the parameters whose variability could be neglected thanks to a very low sensitivity on the results.