Purchase this article with an account.
Trenton Wirth, Gregory Dachner, William Warren; Metric vs. Topological Models of Collective Motion in Human Crowds. Journal of Vision 2018;18(10):1035. doi: https://doi.org/10.1167/18.10.1035.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Collective motion in human crowds is thought to emerge from local interactions between individual pedestrians (Warren, CDPS, in press). A key problem in modeling collective motion is understanding how a pedestrian is influenced by multiple neighbors. Different models have been suggested for modeling the collective motion of animals. A topological model, in which an individual is influenced by a fixed number of neighbors, independent of distance, has been found to describe starling flocks (Ballerini et al., 2008). A metric model, in which an individual is influenced by all neighbors within a zone of fixed radius, has been reported in chimney swifts (Evangelista et al., 2016). Similarly, we have experimentally derived a "soft" metric model, in which neighbor influence decreases exponentially with metric distance, that best describes human crowd behavior (Warren & Dachner, VSS 2017). To test these models we previously manipulated the density of a virtual crowd, and perturbed the walking direction of a subset of neighbors (Warren & Rio, 2014; Wirth & Warren, 2016). Both studies found that the participant's turning response depended on crowd density, contrary to a topological model. In the first study, the perturbed neighbors were in random positions, and the response was stronger in the high density condition than the low density condition. The second study was constructed to elicit a stronger response in the low density condition, by always perturbing the nearest neighbors at fixed distances. Here we use the soft metric model to predict both sets of results. The model closely reproduces the observed effects of density, including the reversal of the high and low density conditions. The results rule out a topological neighborhood in human crowds, and provide support for the soft metric neighborhood model.
Meeting abstract presented at VSS 2018
This PDF is available to Subscribers Only