© Marco Richter/iStock/Thinkstock

Helping trains take the strain

18 Nov 2014

An agent-based model that analyzes commuters’ travel data will improve the Singapore rail experience

A computer model provides valuable predictive data for optimizing the rail commuter experience.

A computer model provides valuable predictive data for optimizing the rail commuter experience.

© Marco Richter/iStock/Thinkstock

The introduction of smartcard ticketing for Singapore’s public transport system has enabled A*STAR researchers to provide valuable predictive data on potential train overloading.  This will enable system planners to address critical bottlenecks as the system stretches to accommodate an expanding population1,2.

Over one million commuters — roughly 20 per cent of Singapore’s population — use the mass rapid transit (MRT) system every day. With the population slated to increase by 26 per cent by 2030, this growth needs to be managed in a way that prevents system delays and overcrowding. A suboptimal transport system could lead to dissatisfied customers and higher economic costs.

To conduct their investigation, Christopher Monterola and colleagues at the A*STAR Institute of High Performance Computing used a modeling technique known as an agent-based model (ABM), which identifies key individual influencers, or ‘agents’, in a complex system and models them in a relatively ‘natural’ way. The team chose three tractable agents: the commuters, the train and the station. Unlike other transportation models, the ABM can consider interactions between agents.

The team examined two main problems that lead to travel delays: overloading and overcrowding. By varying the train’s loading capacity (the maximum number of commuters a train can accommodate at a given time), the team identified a threshold capacity: beyond this tipping point even a few additional commuters produce a cascade of delays. Similarly, more passengers waiting on crowded platforms in popular routes may also significantly increase delays and extend travel times.

Monterola (center right) and colleagues in the Complex Systems Capability Group.

Monterola (center right) and colleagues in the Complex Systems Capability Group.

© 2014 A*STAR Institute of High Performance Computing

Prior to its use for scenario planning, the model was experimentally validated using a week of Singapore smartcard data, which corresponds to 14 million journeys. The data collected for each journey included the anonymized smartcard ID, journey ID, date, origin and destination stations, ‘tap-in’ and ‘tap-out’ times, and the distance traveled.

The model can be used to assist MRT system planners in alleviating strains on a system should it become overloaded through the provision of real-time information on threshold capacities and ‘bottleneck’ stations.

Monterola says his team is passionate about finding ways to improve the robustness and efficiency of the MRT system. “This work is scientifically challenging, but more importantly, it is socially relevant,” he explains.

Other transportation systems could also use the model, which can be “augmented to work with real-time data, to enable a livestream view of all commuter movement in a city,” says Monterola. The team is currently working with behavioral scientists to interpret the influence of these system variables on commuter satisfaction — perhaps ultimately even at the individual level.

The A*STAR-affiliated researchers contributing to this research are from the Institute of High Performance Computing (IHPC). More information about the Complex Systems Capability Group’s research can be found at the group’s page on the IHPC’s website.

Want to stay up-to-date with A*STAR’s breakthroughs? Follow us on Twitter and LinkedIn!


  1. Legara, E. F., Monterola, C., Lee, K. K. & Hung, G. G.  Critical capacity, travel time delays and travel time distribution of rapid mass transit systems.  Physica A 406, 100–106 (2014). | article
  2. Othman, N. B., Legara, E. F., Selvam, V. & Monterola, C.  Simulating congestion dynamics of train rapid transit using smart card data. Procedia Computer Science 29, 1610–1620 (2014). | article

This article was made for A*STAR Research by Nature Research Custom Media, part of Springer Nature