University of Maryland geographer Paul Torrens and his team have created a model to predict the movements of people in crowds, such as during a riot. The simulation incorporates factors such as the desire to avoid collisions, herding and other group dynamics. "We've tested it using data from real people moving around. The model does a good job at replicating human movement," Torrens said. While the simulation probably won't help to stave off riots, he added, it could have applications for emergency planning and disaster relief.