Multiscale Urban Pedestrian Dynamics: Developing More Efficient Predictive Models for Smarter, Safer, Flowing Spaces

Why is it that hundreds of people still die every year, and thousands more are injured, caused by crowd crushes and poorly designed infrastructure? The answer to this question lies in our current inability to model human behaviour accurately, with tragedies continuing to occur even when state-of-the-art crowd control procedures are employed. For example, in 2014 at Edinburgh’s Hogmanay New Year’s Eve festival on Princes Street, 123 people were injured and needed medical attention as a result of congestion on The Mound caused by a famous DJ performing his hit song. The annual pilgrimage in Mecca, which draws several million pilgrims to perform the Hajj within a week, has seen a trail of fatalities over an entire century, with the one in 2015 amounting to more than 2000 deaths due to a single massive crush. These are just two examples of a long worldwide history of deaths that could have been prevented by better research, better understanding, and better predictive software tools.

As urban populations continue to grow rapidly, there is a need to develop more efficient, predictive models to enable future cities to be smarter, safer, flowing spaces. Pedestrian safety is a top priority when designing public infrastructure or when organising public events (e.g. concerts, protests etc.). However, modelling how people move, react and behave in different situations is complex and depends on several factors. Agent-based simulations model people individually and can predict realistic emergent crowd behaviour through underlying rules. However, the models for person attributes used by commercial software, such as social interactions and panic, are often ad hoc, require experimental calibration, and usually break down in many instances.

The objectives of this PhD research project are as follows:
1) To carry out a literature survey on agent-based modelling of crowds;
2) To develop new mathematical and psychological models for accurately and efficiently predicting the underlying behaviour of people in several scenarios;
3) To develop a pedestrian computational framework in the open source OpenFOAM software, and implement and test these new models on High Performance Computing facilities, such as ARCHER (the UK’s national supercomputer based in Edinburgh);
4) To collect experimental data from external sources, and validate the models;
5) To engage with industrial partners and to transfer knowledge;
6) To disseminate the research widely, to the public and decision-makers in urban infrastructure.