Journal of Safety Science and Resilience (Mar 2023)

Study on single-group dynamics in evacuation considering pre-defined and undeclared leadership

  • Wei Xie,
  • Dongli Gao,
  • Ruifeng Cao,
  • Eric Wai Ming Lee,
  • Richard Kwok Kit Yuen,
  • Jingwen Weng

Journal volume & issue
Vol. 4, no. 1
pp. 52 – 60

Abstract

Read online

Grouping is a common phenomenon that occurs everywhere. The leader-follower relationship inside groups has often been qualitatively characterized in previous models using simple heuristics. However, a general method is lacking to quantitatively explain leadership in an evacuating group. To understand the evolution of single-group dynamics throughout an evacuation, we developed an extended social force model integrated with a group force. A series of single-group evacuations from a room were simulated. An information-theoretic method, transfer entropy (TE), was applied to detect predefined and undeclared leadership among evacuees. The results showed that the predefined leader was correctly detected by TE, suggesting its capability in measuring leadership based on time series of evacuees’ movement information (e.g., velocity and acceleration). When evacuees were grouped together, TE was higher than when they were alone. Leaders presented a monotonically increasing cumulative influence curve over the investigated period, whereas followers showed a diminishing tendency. We found that leadership emergence correlated with evacuees’ spatial positions. The individual located in the foremost part of the group was most likely to become a leader of those in the rear, which concurred with the experimental observations. We observed how a large group split into smaller ones with undeclared leadership during evacuation. These observations were quantitatively verified by TE results. This study provides novel insights into quantifying leadership and understanding single-group dynamics during evacuations.

Keywords