While overcrowding on Singapore’s mass rapid transit (MRT) system is predicted to worsen, researchers have combined simulated scenarios with smartcard data to find out how these bottlenecks occur. Abi Millar finds out more from the A*STAR Institute of High Performance Computing.
When Singapore’s mass rapid transit system (MRT) opened in 1988, it was slated as the most practical and efficient travel solution for a densely populated state. Generally seen as the fastest way to get around Singapore aside from taxi, the network has expanded significantly and today forms the backbone of the city’s public transport system.
In recent years, however, the MRT system has faced occasional concerns about overcrowding and disruptions. Travel chaos notoriously grew so bad that crowded trains became a key political point at the 2011 general election.
“I have experienced the discomfort and frustrations that commuters faced because of the congestion and the sometimes unreliable service and I share your desire to see improvements to our public transport,” wrote the Minister for Transport, Lui Tuck Yew, at the time.
While congestion has been alleviated somewhat since then – with operators adding in more than 2,000 train trips a week – there may be further issues lying in store. In the years up to 2030, Singapore’s population is expected to increase by more than a quarter. This will place the transport network under additional strain, and is likely to cause major headaches if the situation is inadequately managed.
However, a possible solution may be in the making, with researchers working on innovative approaches to future problems.
“Singapore is always looking forward, preparing for a range of possible future scenarios,” explains Christopher Monterola of the A*STAR Institute of High Performance Computing. “The concern mainly arises from the question of sustainability and resilience – if and when Singapore’s commuting population, those who are using its rapid transit system, will significantly increase 5 to 10 years down the road.”
What can be done is open to question. In order to find workable answers it will be necessary for developers to determine exactly what they are up against at the moment, and how a surge in passengers might affect the system in the future.
Mining the data
While this is by no means straightforward, a new study by Monterola and colleagues has provided an intriguing starting point. Using a computer modelling technique, they examined how just a small amount of overcrowding might lead to a suboptimal transport system.
“We have built a mechanism-based full-scale agent-based model of the Singapore Rapid Transit System,” he explains. “Using the model, we can look into various scenarios. One of our interesting findings include identifying a critical point in the system at which, if nothing is done about it could lead to a cascade of system failures – commuters not being able to board and/or increase in travel delays.”
The research was made possible thanks to the use of contactless smartcard ticketing across the network. Users simply tap in at one end and tap out at the other, and their anonymised data is recorded (including smartcard ID, journey ID, date, origin station, destination station, tap-in and tap-out times, and the total distance travelled).
Data of this kind is invaluable when it comes to interpreting transport networks, and the introduction of smart ticketing across the world has enabled new insights into how passengers use the services. For instance, in 2011, A*STAR researchers analysed a week’s worth of data at particularly busy points on the MRT. This allowed them to learn more about the dynamic properties of the network, especially the parts that were facing bottlenecks.
For their latest study, the researchers obtained a week’s worth of data across the entire MRT, corresponding to around 14 million journeys from two million unique card IDs. While this information was useful, it was only one piece of the puzzle. Since smart card data does not contain routing information – just the entry and exit points of a journey – it can only tell you so much about how passengers behave in between.
For this reason, the researchers also used a mechanistic agent based model (ABM), which is regarded as a fairly ‘natural’ way of modelling variables in a complex system. Through employing an ABM, computer scientists can simulate how various groups or individuals interact, and how that affects the system as a whole.
“Agent-based modeling (ABM) is a well-established approach in the field of computation – it is especially useful in the study and investigation of various complex systems where there are many interacting parts and whose interactions result to some kind of emergence or self-organisation,” Monterola explains. “The transport system is an example of a complex systems composed of two types of ‘agents’ – the trains and the commuters. Each agent is regulated by certain rules defined in the simulation. A critical component is being able to accurately describe these rules, and in the process make predictions.”
Basing their work on the train dispatch schedule, the researchers also accounted for additional factors such as passengers’ walking times. These simulated travel duration distributions were experimentally validated using the smartcard information – two distinct datasets that luckily matched up well.
“ABM is just one aspect of this work – for example, the rules behind the movements of the commuter agents are shown to be not based on shortest path route, but require some sophisticated statistical physics/thermodynamic model to be well represented,” adds Monterola.
Once the model had been created, the team were able to vary the train’s loading capacity, and the holding power of the platform, to identify the tipping point they were seeking. Before reaching this point, trains ran smoothly, but past this threshold capacity, even a few additional passengers prompted delays.
The researchers concluded that, if the commuter population were to increase by 10% proportionally, the system would reach its limit, with the number of missed trains nearly tripling from 22,000 to 62,000. Past that point, even measures such as expanding peak hours would still prove insufficient, meaning developers would have no choice but to increase train frequency or add new lines.
It is clear there are implications here for the future of the MRT system, indicating that if projected population growth does occur, planners will need to find some strategic ways to accommodate the new passengers. Ultimately, tools of this kind can help them make quantifiable policies that balance out convenience and cost.
Work remains to be done – for instance, the researchers are currently working with behavioural scientists to establish how different variables impact on passenger satisfaction. Still, the utility of the model is clear. In the future, it could be tweaked for use in different cities with similar transportation issues, and even be augmented for use with real-time data. This would make it possible to estimate commuters’ movement in the here and now.
“For our research on the Singapore transport system, the natural flow would be to extend this to other transportation systems,” explains Monterola. “For instance, we have now integrated our model with bus systems in Singapore, and we are now looking into adding traffic simulations in roads due to private cars.’
“We are also looking at first and last mile problem for commuters, and how land use allocations are impacting the spatiotemporal demand dynamics of traffic flow. We are continuously improving our model and adding more features to address the different needs of our collaborators including schedule optimisation constrained by various infrastructure and resources constraints. Overall, our Complex Systems (CxSy) team at IHPC is interested, broadly speaking, in the Science of Cities, Economic Complexity, and Socio-Technical Systems.”
The tool may assume critical importance in the future as problems of congestion become more severe. With limited land space and millions of commuters, Singapore urgently needs to address the overcrowding problems on the MRT system. While this work is scientifically challenging, there can be no doubt that it is a crucial step along the way.
This is the cover feature for the April 2015 edition of Future Rail Magazine