Stopping the stampede: Planning for urban emergencies

Engineers are using mice, ants and virtual reality to study crowd behaviour so infrastructure can cope with attacks and other disasters

Greta Harrison

Published 18 May 2016

Last year’s tragic crowd crush at the annual Hajj near Mecca saw the deaths of over 2000 pilgrims after two large groups converged on a narrow strip of road fortified by immovable barriers.

The disaster highlighted the dangers of dense crowds in confined areas, especially when panic sets in and a stampede develops.

Crowd dynamics is a new field of research, with little reliable data on how crowds behave in situations such as terrorist attacks or natural disasters. As the Earth’s population grows and major safety risks occur in densely populated areas, University of Melbourne researchers are working to prevent crowd crush fatalities and improve the safety of our buildings and transport hubs.

Professor Majid Sarvi, from the School of Engineering’s Centre for Disaster Management and Public Safety, is using virtual reality and crowd modelling with insects and mice to gather information about how crowds react when they’re in an emergency.

Professor Sarvi is studying how crowds behave during emergencies. This video shows a human trial of an evacuation so researchers can understand more about human cognition in exit route choice. Researchers changed the escape layout to see which exits the 150 volunteer evacuees would choose. Video: Professor Majid Sarvi.

The team is analysing Melbourne’s railway stations as a model for how transport users might behave in situations of panic such as terrorist attacks. This information will improve the design of current and future transport infrastructure so that crowds can evacuate quickly and safely.

Professor Sarvi says that like most major cities around the world, Australia’s transport and buildings are designed for normal conditions and not for extreme situations.

“Even the normal conditions we’re experiencing now can cause a lot of problems and near misses. The railway platforms during morning rush are very congested, and with 8 million people projected for Melbourne by 2050, I don’t think we would have the capacity to handle that,” Professor Sarvi says.

A still from the 3D virtual model of Parliament Station in central Melbourne. Researchers will conduct user testing in virtual reality using the Oculus Rift, to study how people would evacuate during an emergency or attack in the virtual station.

Working with Dr Korosh Khoshelham and Dr Mohsen Kalantari, from the Department of Infrastructure Engineering, Professor Sarvi has already built a detailed virtual reality model of one of Australia’s busiest railway stations, Melbourne’s Parliament Station, using a high-resolution 3D laser scan of the station and over 400-high definition photos.

Using an Oculus Rift, the research team will engage users to test various scenarios in the virtual station, including emergency evacuations, fires and smoke.

“In another scenario we create dense conditions so that if you go inside the virtual environment there are a lot of virtual agents around you that limit your vision and understanding of the area,” Professor Sarvi says.

What can ants and mice tell us about crowd panic?

Working with ants, woodlice and mice, Professor Sarvi and his team have been able to establish the optimum design for exits in buildings and infrastructure to enable safe and fast emergency evacuations.

Professor Sarvi says one of the reasons there is not a wealth of data around crowd behaviour in disasters is that you cannot put human subjects through genuine emergency conditions in order to test their responses.

“So we decided to use animals as proxy model. It was quite novel, but the hypothesis was there should be some similarity in group behaviour between animals and humans,” he says.

The research team studied the movement of woodlice in order to understand more about crowd behaviour in emergencies. Picture: Woodlouse/Flickr.

The team used different triggers for the different creatures. Woodlice were exposed to sunlight, and ants to a chemical scent, whereas mice were prompted into action through sharp light and a shaking movement. The groups of animals were then monitored as they escaped from confined spaces.

Professor Sarvi says the experiments show that an exit door in a corner of a space was about 30% more effective than a door in the middle of the space, because there was less conflict with individuals running against each other and less chaotic behaviour.

“We also put a column in front of the exit and this improved the efficiency of the animals getting out. It seems counter intuitive, but when we analysed it, it made sense, because it was breaking down the conflict behaviour,” he says.

“Ants, woodlice and mice are very different in terms of movement, cognition and body size. Ants can’t see, while mice are smart animals with big brains. Woodlice are very slow moving, unintelligent animals with no memory. But still the experiment created the same result.”

Left: With an exit in the centre of the space, the mice in the experiment evacuate in a chaotic manner. Right: When the exit is placed in the far corner, the mice can evacuate quickly, with less conflict of movement. Picture: Professor Majid Sarvi.

Professor Sarvi and his team also ran a series of large-scale human crowd dynamics experiments outside of the virtual realm. Various building floor plans were recreated with large panels and around 150 volunteers were instructed to exit under normal conditions, as well as to evacuate as quickly as possible.

Videos of the scenarios demonstrate that the human trials showed similar results to the studies of insects and mice.

This video shows a human trial of an evacuation in which two crowds merge towards an exit. The objective was to determine the optimum merging angle in an evacuation. The research team also looked at the difference between orderly evacuation (walking) and extreme and emergency evacuation (running). Video: Professor Majid Sarvi.

“We did find some herding behaviour, and some people said that in an emergency they would definitely follow others. That can make a difference in a real emergency situation where the crew on the ground can play a significant role if they know where to go and people can follow them,” he says.

The videos also show how exits in the middle of a space and crossed flows of people moving in different directions creates conflicts and bottlenecks. The team also asked groups of participants to cross paths at different speeds; walking, jogging and running.

“When people walked, they could easily get around each other and there was a smooth flow of movement. When you asked them to jog and run they couldn’t adjust their speed to get around each other and you immediately saw some bottlenecks.”

Professor Sarvi says infrastructure designers need to avoid instances in which the flow of people could come into conflict.

The merging and crossing of high density crowds is bad design.

“You need to also have a good understanding of how you can create an escape route for people if something goes wrong,” he says.

Professor Sarvi is continuing to work on the virtual reality trials with Public Transport Victoria and Victoria’s Department of Economic Development, Jobs, Transport and Resources. He says his team is also hoping to expand the research to include MRI trials that would provide insight into human brain activity during crowd panic.

Banner image: Pixabay

Find out more about research in this faculty

Engineering & Technology

Content Card Slider

Content Card Slider

Subscribe for your weekly email digest

By subscribing, you agree to our

Acknowledgement of country

We acknowledge Aboriginal and Torres Strait Islander people as the Traditional Owners of the unceded lands on which we work, learn and live. We pay respect to Elders past, present and future, and acknowledge the importance of Indigenous knowledge in the Academy.

Read about our Indigenous priorities
Phone: 13 MELB (13 6352) | International: +61 3 9035 5511The University of Melbourne ABN: 84 002 705 224CRICOS Provider Code: 00116K (visa information)