Mesoscopic Modeling of Train Operations: Application to the MBTA Red Line

Saeid Saidi1, Nigel Henry Moir Wilson2, Haris N. Koutsopoulos3, Jinhua Zhao2

  • 1Massachusetts Institute of Technology
  • 2MIT
  • 3Northeastern University


11:45 - 12:00 | Mon 28 October | The Great Room III | MoC-T4.4

Session: Special Session on Smart Railways (I)

Category: Special Session


In this paper, we introduce a train following model which can efficiently capture the effects of train interactions in an urban rail line. The train following model is based on the estimation of induced delay for a train closely following a lead train, and is derived from empirical analysis based on historical track circuit data. Based on an analysis of sequential train delays, a mesoscopic train state prediction model is developed which can be used to predict the behavior of the system with respect to changes in initial conditions (e.g. scheduled headway, headway variation or dwell time) and disruptions. The performance analysis using this train following model is richer and more accurate than that from analytical macroscopic models while not being as time- and resource- intensive as a detailed simulation model. The development and potential application of this model is demonstrated for the Massachusetts Bay Transportation Authority (MBTA) Red Line. The model can be used for both real-time control strategy evaluation and offline operational strategy analyses.