Statistical features of the passage of a pedestrian crowd through a bottleneck

ALEXANDRE NICOLAS; BOUZAT, SEBASTIAN; MARCELO KUPERMAN

Zaragoza

Conferencia; BIFI International Conference 2018; 2018

Universidad de Zaragoza

The motion of a pedestrian often looks erratic when it is examined individually. Investigations conducted in the last fifteenyears have however revealed that the dynamics of pedestrian crowds exhibits interesting statistical features. This is especiallytrue for flows through a bottleneck, a problem that is central for the optimisation of flows and the design of facilities that canbe evacuated quickly. Controlled evacuations through a narrow door conducted by Zuriguel et al. (Scientific Reports 4, 2014)have shown that the microscopic escape dynamics can be robustly characterised, with statistical laws strongly reminiscent ofgranular flows through an orifice. In the wake of this work, we experimentally studied the influence of the individual behaviourson the dynamics of evacuation, by prescribing either a selfish behaviour or a polite one to participants (Nicolas et al., Transp.Research B 99, 2017) . An original setup allowed us to conduct the analysis is in a macroscopically quasi-stationary regime.Macroscopically, in the regime of limited crowd pressure, we observe that several flow properties are insensitive to thebehavioural details, but rather depend on the local pedestrian density at the door, at least for limited crowd pressure. Inparticular, the global flow rate grows monotonically with the pedestrian density, under these conditions. Microscopically, thefollowing statistical laws are inferred : (i) In competitive settings, the distribution of time gaps between successive escapesdisplays a heavy (power-law-like) tail, hinting at long-lived clogging events ; (ii) surprisingly, these time gaps are anticorrelatedin time, pointing to an alternation between shorter time intervals and longer ones ; (iii) bursts of almost uninterrupted escapes(i.e., separated by small time gaps) are exponentially distributed. We develop and exploit simple quantitative models to shedlight on the mechanisms at the origin of these statistical features, whose generality is thus underlined.