Evaluating crowding in individual train cars using a dynamic transit assignment model

As travel demand grows in many cities around the world, overcrowding in public transport systems has become a major issue and has many negative effects for both users and operators. Measures to address on-board congestion span from large-scale strategic investments (e.g. increasing infrastructure capacity), through tactical planning (e.g. stopping pattern) to real-time operational measures (e.g. information provision, gate and escalator control). Thus there is a need to evaluate the impact of these measures prior to their implementation. To this end, this study aims at capturing the effective capacity utilization of the train, by considering passengers' distribution among individual train cars into an agent-based simulation model. The developed model is validated and applied to a case study for the Stockholm metro network. The findings suggest that an increase in peak hour demand leads to a more even passenger distribution among individual train cars, which partially counteracts the increased disutility caused by the higher passenger volumes. Interestingly, the closure of the most popular entrance point at one of the stations leads to lower train crowding unevenness at the downstream stops and consequently reduces passengers' experienced discomfort. We find that the user cost is significantly underestimated when passenger distribution among cars is not accounted for.