Published online 28 September 2000 | Nature | doi:10.1038/news000928-9


Panic over?

Panicking people are unpredictable -- or are they? Philip Ball investigates a potentially life-saving computer model.

How does an orderly crowd turn into an panicked stampede? A new computer model may hold the answer.How does an orderly crowd turn into an panicked stampede? A new computer model may hold the answer.

Engineers and architects try to design public spaces and buildings expecting the worst; but the worst is extremely hard to predict. When panic strikes, people do not behave rationally -- they do not calmly seek the nearest exit, nor do they file out in orderly fashion.

New research now offers a chilling insight into how panicking crowds do behave. Physicists from Hungary and Germany have devised a model they claim will allow more accurate forecasting of behaviour during emergencies. They report the first results in Nature1.

Many people have experienced the awful slide from congested but courteous crowd motion into uncontrolled panic. People are trampled, steel barriers are bent -- and the pressure of a crowd has even been known to push down brick walls. But it is difficult to provide a clear, general description of that irrational transition, especially one grounded in physics.

Yet this is what Dirk Helbing, of the Dresden University of Technology, and colleagues have done. In a crowd, people tend to want to move in a particular direction with a specific speed and try to avoid collisions. This can be modelled using a computer: the team ran simulations in which blob-like 'people' tried to exit a room through a single door.

Under normal conditions a steady flow of people pass through the door and a dense, but relatively orderly, mass of people cluster around it. But when the average desired velocity of the crowd increases above a critical point, there is pandemonium.

Quite suddenly, the computerized crowd switches from being well behaved to frantic. The desire for speed overwhelms the desire to avoid collision and the blob people jam up against one another -- just as salt can jam the shaker even though the hole is bigger than the largest grain. The room takes longer to empty even though everyone tries to move faster -- handfuls of people escape in bursts between clogging events.

This classic case of 'more haste, less speed' shows that calculations of room-emptying times based on orderly behaviour poorly predict what happens in a panic situation.

The researchers also investigated how a crowd would find its way out when the exits could not be seen (in a smoke-filled room for instance). In this situation people balance a tendency to follow others -- a 'herd instinct' -- against an tendency to find their own escape route.

The simulation reveals that the best escape plan in this situation is a mixture of herding behaviour, which allows many people to escape when they find a door (but can lead to jamming) and individualism.

But in many cases the computerized exits are not used as intended: most people crowd around one door, while another is almost ignored. So here is another way that conventional calculations of exit rates can get things wrong -- and how the new model might help to get them right. 

  • References

    1. Helbing, D., Farkas, I. & Vicsek, T. Simulating dynamical features of escape panic. Nature 407, 487 - 490 2000. | Article | PubMed | ISI | ChemPort |