[Paper Review] Microsimulations of Arching, Clogging, and Bursty Exit Phenomena in Crowd Dynamics

Francisco Enrique Vicente Castro, Jaderick P. Pabico|arXiv (Cornell University)|Jun 25, 2015

Evacuation and Crowd Dynamics8 references3 citations

TL;DR

This paper presents a microsimulation model of crowd dynamics using artificial agents based on social comparison theory and field-of-vision-aware trajectory mapping, successfully replicating real-world phenomena such as arching, clogging, and bursty exit rates. The model reveals a novel emergent behavior—double arching—occurring during temporary calm phases in exit density, enhancing realism for evacuation and crowd management simulations.

ABSTRACT

We present in this paper the behavior of an artificial agent who is a member of a crowd. The behavior is based on the social comparison theory, as well as the trajectory mapping towards an agent's goal considering the agent's field of vision. The crowd of artificial agents were able to exhibit arching, clogging, and bursty exit rates. We were also able to observe a new phenomenon we called double arching, which happens towards the end of the simulation, and whose onset is exhibited by a "calm" density graph within the exit passage. The density graph is usually bursty at this area. Because of these exhibited phenomena, we can use these agents with high confidence to perform microsimulation studies for modeling the behavior of humans and objects in very realistic ways.

Motivation & Objective

• To develop a realistic agent-based model of pedestrian crowd behavior that replicates emergent phenomena observed in real-world evacuations.
• To investigate how social comparison and vision-based goal navigation influence collective crowd dynamics.
• To identify and characterize new emergent behaviors such as double arching in crowd simulations.
• To analyze the impact of exit width on arching formation and egress efficiency.
• To validate the model’s realism by reproducing key phenomena like clogging and bursty exit rates.

Proposed method

• Agents are modeled using social comparison theory to guide decision-making in competitive environments.
• Trajectory mapping is implemented based on each agent’s field of vision to determine movement paths toward goals.
• The simulation environment features a corridor with a narrow exit, allowing observation of spatial competition and flow dynamics.
• Density graphs are used to monitor and analyze exit rate fluctuations and clogging events.
• Varying exit widths are tested to study their effect on arching profile dimensions (major and minor axes).
• Microsimulations are conducted to observe emergent patterns such as arching, clogging, and double arching over time.

Experimental results

Research questions

• RQ1 How do social comparison and vision-based navigation contribute to the emergence of arching in pedestrian crowds?
• RQ2 What causes bursty exit rates in crowd simulations, and how do they correlate with clogging?
• RQ3 Can a new phenomenon—double arching—be observed and characterized in simulation?
• RQ4 How does exit width influence the shape and size of the arching formation?
• RQ5 To what extent does the proposed agent model replicate real-world crowd dynamics compared to existing models like SFM?

Key findings

• The model successfully reproduces arching, clogging, and bursty exit rates, confirming its realism in simulating crowd dynamics.
• Double arching emerges during temporary calm phases in the exit density graph, representing a novel emergent behavior not previously documented.
• Clogging is observed when two or more agents compete for space near the exit, leading to flow disruptions.
• Bursty exit rates are characterized by a saw-tooth-like profile in the density graph, indicating intermittent flow.
• Narrower exit widths result in longer major axes and shorter minor axes in the arching profile, while wider exits produce the opposite effect.
• The model demonstrates potential for simulating complex behaviors such as imitation and contagion, which are absent in the social force model.