Oasys Software Blog

Read a recent article in Bloomsberg Businessweek which features MassMotion’s lead developer Erin Morrow.

Bloomsberg Businessweek

http://www.businessweek.com/magazine/erin-morrows-massmotion-simulates-crowds-11172011.html

There has been a suggestion in some quarters that crowds behave like fluids. This may be a useful paradigm when the entire crowd is moving in the same direction down a pipe, sorry, tunnel, but it is rather unrealistic when you examine pedestrian contra flows: you cannot have water moving down a pipe in both directions at once. Fluids, by definition, will also expand to fill the available space, which crowds do not.

In reality, crowds are made up of individuals and small groups that each have their own agendas and destinations. What does happen however in crowds moving in opposite directions is that lanes spontaneously form: people find it easier to follow in the wake of someone going in the same direction.

Having seen some online videos of crowd simulations based on fluid principles, I wondered how a more realistic simulation would deal with the problem of eight crowds simultaneously trying to move from one side of a space to another, a bit like a giant version of Chinese Checkers.

In the first video, the MassMotion pedestrian agents are all set to move at the same speed and you thus get a moderate amount of rotation in the crowd due to the keep-right preference, but is coupled with a steady evaporation around the edges until the central mass reduces to too small a size to be sustained.

In the second video the agents are given a more usual variation in speeds, leading to a more dynamic integration in the mass and what looks rather like turbulence as lanes form.

Helbing in his 2001 paper Traffic and related self-driven many-particle systems (Reviews of Modern Physics, 73, 1067-1141) observed that “roundabout” flow patterns emerge at intersections. If we model such an intersection in MassMotion then this is exactly what we see in the analysis:

Intersections

This “coriolis effect” indicates that while it is not realistic to model crowds as fluids, there is some fluid-like behaviour that emerges from the individual agents’ movements.