Sunday, November 20, 2011

Virtual rioters predict how crowds move

 
How did recent UK riots spread so quickly to a larger area? Although there are several phenomena at work, a new simulation, developed by geographer Paul Torrens from University of Maryland and his team, is helping to visualise how groups of people tend to move around. "Whereas many existing approaches use particle physics, my approach is sourced in theory and observation," says Torrens.

Based on how people interact with their environment, the model incorporates about 30 different behaviours, from processes like path planning and collision avoidance to group dynamics such as herding and following a leader. Concepts are translated into mathematical models by applying statistics or using methods like game theory. Different algorithms can be applied to each agent, or to the group as a whole, to see how they affect overall behaviour.

The video begins with a simulated riot where characters are static and where the emphasis is on social interaction, decision-making and the connection between geography and social agency.  But the crowd models that follow focus on how a group of people moves around. They can also be experienced as immersive graphical environments on a variety of platforms, such as Xbox or Kinect.

According to Torrens, the simulation is unlikely to help prevent riots due to their complexity. But it's a realistic representation of mass behaviour in terms of geographical factors. "We've tested it using data from real people moving around," he says. "The model does a good job at replicating human movement."

With some improvements, the system could be used to help with disaster relief. Torrens plans to refine the way people interact with each other and how they respond to changes in their environment. "We're building a dynamic environment where buildings collapse and agents plan their behaviour in response, for example by avoiding falling objects," he says. "We plan to develop an Xbox version of this system for use in emergency planning."  

If you enjoyed this video, you might like to see 3D city models produced from crowd-sourced photos or check out a realistic virtual crowd.

http://www.newscientist.com/blogs/nstv/

No comments:

Post a Comment