Abstract
Collective motion in human crowds is thought to emerge from local interactions between individual pedestrians. A key problem in understanding these interactions is the structure of a pedestrian's visual neighborhood, that is, which neighbors influence the pedestrian's behavior. Several hypotheses have been proposed (Strandburg-Peshkin, et al. 2013). The metric hypothesis (Reynolds, 1987; Helbing & Molnar, 1995) supposes that a pedestrian is coupled to all neighbors within a fixed distance, or metric radius. The topological hypothesis (Ballerini, et al., 2008) supposes that a pedestrian is coupled to a fixed number of nearest neighbors, regardless of their distance. We dissociate these hypotheses by manipulating the density of neighbors in a virtual crowd. The metric hypothesis predicts that a pedestrian's response should depend on density, whereas the topological hypothesis predicts it should not. Participants wore a head-mounted display and walked in a crowd of virtual humans, while head trajectory was recorded. During the trial, a subset of the nearest neighbors (S=0, 2, 4, 6) turned 10° left or right, and participants were instructed to "walk with" the crowd. In Experiment 1 (N=11), the 6 nearest neighbors appeared at constant distances (.25m apart in depth), while the density of the remaining 18 neighbors was varied (.25m, .50m apart in depth). Mean final heading direction increased linearly with subset S (p< .001), but more steeply in the low density than the high density condition (p< .035 at S=6), as predicted by the metric hypothesis. In Experiment 2, we test 12 neighbors at more extreme densities that predict a greater difference in final heading (4, 8 or 12 within estimated radius of 4 m). The results are contrary to the topological hypothesis, but are consistent with a metric visual neighborhood. Related findings suggest a flexible metric radius that may depend on the distance of the nearest neighbor.
Meeting abstract presented at VSS 2016