Virtual Crowds That 'See' Can Help Predict Real Disasters
Stampedes such as one that killed 19 concertgoers at a "Love Parade" event in Germany last summer could be avoided if planners could predict how crowds move. A new crowd simulation gets better results than past physics-based models by giving the virtual people the ability to "see" as they navigate pedestrian traffic.
European researchers found that the vision-based approach predicted behaviors very well for both large and small crowds, Technology Review reports. Their tool could prove especially useful for analyzing crowd disasters that take place in smoky or limited-vision conditions, as well as improving how robots deal with human crowds.
Older models deal with crowd behavior in a simplistic way by treating people like particles bouncing off one another. But that has proved unsatisfying for predicting real-life behavior in many cases, and so researchers have begun looking for new approaches. [Predicting the Next Egypt: The Recipe for Revolution]
The vision-based simulation can shortcut the dilemma of trying to figure out how to simulate complex thinking processes for each person. Seeing where you're going affects both walking direction and speed in crowds, and so the new model came up with behavior predictions that matched almost perfectly with real-life crowd scenarios.
Still, people can't always move where they wish in a rowdy or possibly panicked crowd. Researchers also modeled the physical contact of the crowd in their simulation, so that they could predict what might happen in the tight quarters that can lead to crowd disasters.
Predicting the behavior of virtual agents based on supposed sight and physical contact goes beyond life-or-death scenarios. Even some games such as the popular "Assassin's Creed" series have featured basic crowd behaviors, complete with shoving and groups gathering around dead bodies or stopping to pick up money strewn across the ground.
The study is detailed in the April 18 issue of the journal PNAS.
This article was provided by InnovationNewsDaily, a sister site of TechNewsDaily.
- Tech Tracks Social Media Crowd Reactions to TV
- Predicting the Next Egypt: The Recipe for Revolution
- New Gamer's Strategy: Stoop to Artificial Intelligence