Can we predict panic in crowds by watching them move?

Researchers combined kinetic theory (treating crowds like gases) with ensemble data assimilation to track pedestrian density and estimate a "panic factor" in real time. Testing the approach against agent-based simulations—which use different rules—showed the method still works: observations of actual movement patterns forced the model back on track despite structural mismatches. This suggests crowd models need constant data feedback to stay accurate; predictions without measurements diverge from reality.