Understanding how people move about within towns and cities could offer huge benefits for improving services such as crime prevention, or traffic pollution control. But building a dynamic model that can provide that kind of data in real time poses a big challenge.
Census and other types of household survey readily yield a wealth of information about where people live, but finding out about where they work and how or when they travel is more difficult. That’s because current data collection is patchy and because people’s routines are fluid, particularly outside rush hour and other peak times.
The SURF project (Simulating Urban Flows) is tackling this by harnessing new developments in big data collection and analysis. Using information from social media, mobile phone usage and other sources, LIDA researchers are building an ‘agent-based model’ – a computer programme that can simulate the effects of individuals on a whole system.
This type of system is incredibly complex to produce, so the team is working initially on a more contained project, focusing on commuter behaviour in the town centre of Otley, just outside Leeds. This was selected because it’s a reasonably small area, with access to good data sources. Post-doctoral researcher Tomas Crols, who is working on building the model, is drawing together the most useful sources, in particular data from Noggin, a company which uses sensors placed in shops to count when a mobile phone passes by.
Dr Nick Malleson, leading the project, explains: “Data sets from organisations like Noggin give us a really good idea of how many people are walking past certain points at different times of the day. We don’t track individuals, only the counts of people, and inputting this data across different times helps us build an increasingly accurate picture of footfall throughout the day.”
“Once we’re confident we have a reasonably accurate model of what’s going on amongst commuters in Otley town centre on a typical weekday, we’ll start to diversify the model, looking at shoppers, for example, or retired people. This will be a real challenge – one way to do this could be to run lots of models, with different types of individuals in them, and test them to see which simulate the real world the most accurately.
As the team builds the complexity and sophistication of the model, they can also start to think about broadening the geographical area within which they are working, incorporating more diverse modes of transport, and bringing in partners who might be interested in using the model.
“The insights we’ll be able to provide might be useful to local authorities who want a more accurate picture of which streets are worst for pollution at particular times of day, taking the movements of traffic and of pedestrians into account – it’s the kind of data that could be used to inform policy on environmental improvements,” says Dr Malleson. “Law enforcement agencies can also use it to gain more sophisticated overviews of where crime hotspots are, based on where groups of potential crime victims are at different times of the day, rather than simply where they live.”
In a related project, funded by the European Research Council, the team is also investigating ways to introduce ‘live’ real-time data, creating a system similar to that used in weather forecasting, where the model is continually updated to provide the most accurate and up-to-date picture possible.
Called the Dynamic Urban Simulation Technique (DUST), the model is being developed to help agencies respond to civil emergencies, such as flooding, terrorist attacks or fire.
“Making decisions about how to respond effectively to rapidly-changing situations can be extremely difficult,” says Dr Malleson. “Current agent-based models rely on historical data, so don’t reflect what is happening ‘on the ground’. The model we’re aiming to build through this project will assimilate ‘live’ data, so can be used as a tool to understand and react to situations in real time.”