Joint Macro- and Microscopic Modeling of Pedestrian Flows in Train Stations, with Focus on Train Platforms
Bachar Kabalan, Ecole Des Ponts ParisTech, Mohamed Jaouad Lahgazi,, Ecole des Ponts ParisTech, Fabien Leurent, Ecole des Ponts ParisTech
The paper’s objective is to provide a joint modeling of pedestrian flows in a railway station, by combining macroscopic simulation and microsimulation.
In big metropolitan areas, the largest capacities for passenger transportation are provided by railway lines, with up to 60 or 100 thousands of passengers per hour and per track on the busiest sections (e.g. in Hong-Kong, Paris etc.). Transit flows of such magnitude involve related massive passenger flows in the train stations along the line, be it for boarding, alighting, transferring between lines, or station entry and exit. Thus, every large railway station in a metropolitan area constitutes an important traffic system, which interacts with the traffic systems of the incoming railway lines as well as with local transportation around the station.
The station traffic system is very sensitive to disruptions in train traffic, since a delayed train will carry more passengers, among them more alighting people, and will be awaited for more boarding passengers. This makes the train platform a critical point for passenger flowing. High on-platform passenger densities call for much attention from the station operator, since they induce the risk of passenger falling from platform to track – a twofold risk of jeopardizing the passenger’s life and blocking train traffic along the line during a long slice of time.
This is why a stream of research has recently thrived on passenger traffic in railway station. Two lines of attack have been taken: first, the microsimulation of pedestrian traffic [1,2], second, the inference of passenger densities across the station on the basis of a variety of traffic sensors [3,4].
The paper’s objective is to provide a joint modeling of pedestrian flows in a railway station, by combining macroscopic simulation and microsimulation. Macroscopic simulation is suited to deal with very large flows involving up to several dozen thousand individual passengers, whereas microsimulation enables one to finely model traffic situations involving dozens or hundreds of passengers together with platform features (including length and width, layout and furniture etc.) and train characteristics (such as total length, number of cabins, number and width of doors, alighting flows and dwelling times).
The ultimate goal is to model the dynamic traffic state of the railway station and identify the main points of congestion. To this end, the LADTA (macroscopic) and CapFlow (microscopic) models are used. The original LADTA model for dynamic roadway traffic assignment has been adapted to pedestrian flows and train traffic: the resulting model takes into consideration travelers’ characteristics and route choice behavior, traffic bottlenecks on network links, vertical queuing, and the dynamic capacities for boarding / alighting for every train service. The input data are obtained from field observations, train timetable, and smart card data analysis. Then, the passenger loading of the platform and the evolution of congestion on the bottlenecks (stairs, escalators, and train doors) are quantified. A selection of the model results are then transferred to the microscopic level.
The CapFlow model of crowd dynamics is a 2D discrete pedestrian model, based on non-smooth mechanics and on the application of the Discrete Element Method (DEM) to granular media. This model is able to simulate, with different techniques, the two main pedestrian-pedestrian and pedestrian-obstacle interactions, which are collision and avoidance. The input data consist in the flow values that were obtained by LADTA on the stairs and escalators leading to the platform. The model then simulates the pedestrian dynamics on these elements and on the platform. By visualizing the dynamic interaction of individual passengers and their environment, it is possible to grasp the reasons behind congestion and its dynamics.
After introducing the two models, we apply them jointly to the BFM station (BFM stands for Bibliothèque François Mitterrand) that connects RER line C (RER for Regional Express Railways) to Metro line 4 and Tramway line 3 in Central Paris. The respective results are compared and their consistency is checked, with special focus on the pedestrian flows on train doors and the on-platform stock of passengers after a train departure.
 Galiza, R.J., Kim, I., Ferreira, L., & Laufer, J., (2009), Modeling pedestrian circulation in rail transit stations using micro-simulation, Proceedings of the 32nd Australasian Transport Research Forum (ATRF).
 Hoogendoorn, S.P. & Bovy, P.H.L. (2004), Design assessment of Lisbon transfer stations
using microscopic pedestrian simulation, WIT Transactions on The Built Environment.
 Hoogendoorn, S.P. & Bovy, P.H.L. (2000), Gas-Kinetic Modeling and Simulation
of Pedestrian Flows, Transportation Research Record, No. 1710, pp 28-36.
 Hänseler, F.S., Bierlaire, M., & Scarinci, R. (2016), Assessing the usage and level-of-service of pedestrian facilities in train stations: A Swiss case study, Transportation Research Part A Policy and Practice, No. 89, pp 106-123.
Association for European Transport