September 2021
Volume 21, Issue 9
Open Access
Vision Sciences Society Annual Meeting Abstract  |   September 2021
The neighborhood of interaction in human crowds is explained by visual information
Author Affiliations & Notes
  • William Warren
    Brown University
  • Gregory Dachner
    Brown University
  • Footnotes
    Acknowledgements  NIH R01EY029745, NSF BCS-1431406, NIH T32-EY018080-08
Journal of Vision September 2021, Vol.21, 2840. doi:
  • Views
  • Share
  • Tools
    • Alerts
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      William Warren, Gregory Dachner; The neighborhood of interaction in human crowds is explained by visual information. Journal of Vision 2021;21(9):2840.

      Download citation file:

      © ARVO (1962-2015); The Authors (2016-present)

  • Supplements

Most models of collective motion are based on the physical positions and velocities of others in a neighborhood of interaction. For example, in our physical model (PRSB 2018, CDPS 2018) a pedestrian matches the average heading direction and speed of neighbors, weighted by their distance: weights decay gradually to the nearest neighbor, and more rapidly in the crowd. Recently we developed a visual model, in which a pedestrian’s heading and speed are controlled by nulling the average angular velocity and optical expansion of neighbors, which are sinusoidal functions of eccentricity (VSS 2017, 2019). Neighbor influence is proportionally reduced by partial occlusion (ICPA 2019). Here we use these models to simulate data from three experiments with no free parameters. Exp. 1: Participants “walked with” a virtual crowd of 12 neighbors. A subset of neighbors (0, 3, 6, 9, or 12) changed direction (±10˚) or speed (±0.3m/s) on each trial, and the participant’s walking speed and heading were recorded. The RMSE of heading was 1.97˚ for the visual and 2.08˚ for the physical model (BF=1.90, anecdotal evidence for visual). The RMSE of speed was 0.063m/s and 0.064m/s, respectively (BF=2.43, anecdotal for visual). Exp. 2: The distance of the virtual crowd was varied (2, 4, or 6m), and one row (near, middle or far) changed direction (±10˚). The RMSE of heading was 2.5˚ and 3.6˚, respectively (BF>>100, decisive for visual). Exp. 3: A human ‘swarm’ (N=10, 16, 20) walked together for 2min trials. We simulated thirty 10s segments, modeling one participant with input from neighbors. The RMSE of heading was 15˚ and 29˚, respectively (BF=6, substantial for visual). Remarkably, the gradual decay with distance is explained by Euclid’s law, while the rapid decay is an additional effect of occlusion. The neighborhood of interaction is thus explained by visual information, eliminating explicit distance terms.


This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.