Title:Can high-density human collective motion be forecasted by spatiotemporal fluctuations?

Abstract:The aesthetics of collective motion - be it the swooping murmurations of starling flocks, streams of schooling fish, or turbulent mayhem in mosh pits - compliment the technical advances made by the science of collective behavior. Mathematical models at the interface of fields ranging from evolutionary biology to statistical physics have substantially advanced our understanding of how large-scale coordination emerges from local interactions. Currently, the grand challenge in the field of collective behavior is to develop new empirical methodologies for acquiring larger-scale higher-quality data sets. In this work, we offer a new challenge for the field: real-time forecasting of collective motion. We tackle this challenge in the context of high-density human crowds found at mass gatherings. In particular, we use mode analysis methods from material science and concepts from non-equilibrium physics to study a video recording of a human crowd at a rock concert. With these methods, we analyze positional fluctuations during a period of crowd turbulence to predict the spatial patterns of an emergent human density wave. We focus on these collective motions because they are generally known to cause injuries in extreme circumstances. In addition to predicting spatial patterns of collective motion, we also identify and measure temporal patterns that precede the density wave and forecast its appearance by $\approx$ 1 s. Widening this forecasting window beyond 1 s will require analysis of larger-scale higher-quality crowd data, and likely lead to new computer vision technologies for real-time crowd monitoring and risk-assessment. Such advances would result in reduced rates of injury and loss of life at rock concerts, sporting events, religious pilgrimages, and other mass gatherings.