- Course: CDT for Data Analytics and Society
- PhD title: Developing methods for real-time pedestrian simulation
- Nationality: British
Keiran Suchak is studying for a Centre for Doctoral Training (CDT) programme in Data Analytics and Society, alongside Leeds City Council. He is supervised by Dr Nicolas Mallesonand Dr Jon Ward, and is based in the Leeds Institute for Data Analytics (LIDA). He also collaborates Dr Minh Kieu.
“Many of the PhD projects run by the CDT for Data Analytics and Society are run in conjunction with external partners; in my case, my project is run with Leeds City Council,” Keiran said.
He added: “Leeds City Council is the local authority for the City of Leeds, and is responsible for the provision of education, social services and much more. The Council’s interest in the project stems from the potential benefits that may be achieved using real-time pedestrian simulation methods.”
The Council’s interest in the project stems from the potential benefits that may be achieved using real-time pedestrian simulation methods.
To find out more about CDT training programmes and other PhD study options, visit our research degrees pages.
Real-time simulation of pedestrians
Keiran explained his research involves combining agent-based models with Big Data. The purpose is to facilitate real-time predictions of the population’s movements.
He said: “My research aims to progress the development of real-time pedestrian simulation methods. One of the most popular approaches for simulating pedestrian motion is using agent-based models.
“Agent-based models simulate systems at a microscopic level, characterising individual people as agents with their own autonomy; agents interact with each other and with the environment around them with their own behaviours, resulting in the emergence of macroscopic phenomena such as crowding or lane formation.”
Keiran continued: “Such models are typically calibrated using historical data before being run in an attempt to ensure that they are reflective of the real system. Ultimately, however, these models usually incorporate some degree of randomness to emulate the variability of human behaviour which therefore result in the model diverging from the real system.”
One of the solutions to this problem, Keiran explained, is the use of Big Data. Data are being generated in increasing volumes at an increasing velocity, and data regarding pedestrian flows are no exception, he explained.
Keiran said: “Leeds City Council periodically collects data regarding the number of people in different parts of the city, which could be used in conjunction with an agent-based model of pedestrian motion to simulate the flow of pedestrians around the city.
Leeds City Council collects data... which could be used in conjunction with an agent-based model... to simulate the flow of pedestrians around the city.
“The issue with data, however, is that it is spatially and temporally sparse - in this case, observations are collected at eight locations around the city on an hourly basis. We would, therefore, like to use the data together with the model to gain an understanding of what is happening in the city between these points in space and time.”
At present, Keiran explained, there is not an agreed-upon strategy for incorporating new observations into the model while is it running. This makes it difficult to reliably use agent-based models to simulate pedestrian systems in real-time.
“The incorporation of observations into running models is a problem that is frequently tackled in fields such as numerical weather prediction. Meteorologists make use of data assimilation techniques to introduce observations into their models, updating their simulations,” said Keiran.
“My work, therefore, focuses on adapting one of these data assimilation techniques - specifically the Ensemble Kalman Filter - to work with an agent-based model of pedestrian motion.”
Conducting original research
Keiran investigates a series of questions about the novel use of an advanced data assimilation technique as part of his research. This includes exploring which conditions are needed to improve the accuracy of the model, and how this simulates the system of people. It also involves investigating how the technique will perform when handling observations in different forms, and whether it can be used to infer the ultimate destination of pedestrians in the system.
“First and foremost, my research has involved developing a Python codebase to implement the Ensemble Kalman Filter such that it can interface with a pedestrian agent-based model that has been developed in the research group,” Keiran said.
“Having achieved this primary goal, my research now focuses on a deeper exploration of the effectiveness of the technique, asking wider questions.”
Keiran continued: “At present, the majority of pedestrian modelling takes place with historical data and therefore does not make use of new real-time data as it becomes available. In many cases, this is not an issue of data scarcity but of a lack of appropriate methods for updating models with real-time data.”
Working with Leeds City Council
“The implementation of policy based on such models therefore also fails to take account of data as they become available,” said Keiran. He explained how incorporating the data assimilation technique in a novel way could offer advantages.
Keiran continued: “Leeds City Council holds an abundance of data regarding pedestrian flows through the city centre, and provide said data towards the project. The development of methods to leverage this data in close to real-time would enable the Council to more effectively provide services to the public, as well as enabling the real-time response of adaptive public evacuation strategies.”
The development of methods to leverage this data in close to real-time would enable the Council to more effectively provide services... and the real-time response of adaptive public evacuation strategies.”
“Beyond the ongoing relationship and the data provided, I have also completed an internship with Leeds City Council as part of the CDT program. This focused on exploring trends in the number of people visiting Kirkgate Market and attempting to build models to estimate how many people would visit the market if it were open on Sundays.”
I have also completed an internship with Leeds City Council as part of the CDT program. This focused on exploring trends in the number of people visiting Kirkgate Market...
Deciding to study a PhD
“My first encounter with this field of research occurred when I was doing my MSc Mathematics,” Keiran said.
“As part of this course, I undertook my dissertation with Dr Jon Ward focussing on methods for modelling dynamics on networks. While working on this project, I also attended seminars run by the Leeds Applied Non-linear Dynamics group at which Jon presented a paper he had recently published.
“This paper describes how data assimilation methods (commonly used in numerical weather prediction) may be applied to agent-based models of pedestrian systems – an approach that has since grown in popularity in the field of urban analytics.”
Keiran added: “This was my first real insight into how the mathematical and computational skills that I had picked up over my academic career could be applied to the study of social phenomena.
“Furthermore, it drew together methods from numerical weather prediction which I had been exposed to whilst previously undertaking an internship at the Met Office, and stochastic simulation techniques which I had applied when completing both my undergraduate Physics degree and my Mathematics Masters.”
After finishing his MSc, Keiran moved to London to work as a software developer and analyst, but maintained contact with his supervisor Jon. He explained how he decided to successfully pursue a PhD.
“Over the year, I became aware that the CDT was being formed, and that he would be co-supervising a project which would follow on from the work presented in the aforementioned paper as part of the program,” Keiran said.
I decided to pursue the CDT knowing I would be able to apply my computational and mathematical skills to a project that was at the cutting edge of urban analytics.
“I decided, therefore, to apply in the hope that I would be able to apply my computational and mathematical skills to a project that was at the cutting edge of urban analytics.”