Social Force Model parameter testing and optimization using a high stress real-life situation

TL;DR

This study analyzes a real-life emergency-like evacuation during a Black Friday event, compares it with Social Force Model simulations, and optimizes key parameters to better match empirical data.

Contribution

It provides empirical measurements from a high-stress evacuation scenario and optimizes SFM parameters to improve simulation accuracy.

Findings

Empirical pedestrian flow was higher than laboratory conditions.

Laboratory-calibrated parameters lead to faster simulated evacuations.

Optimized parameters successfully replicate empirical evacuation behavior.

Abstract

The escape panic version of the Social Force Model (SFM) is a suitable model for describing emergency evacuations. In this research, we analyze a real-life video, recorded at the opening of a store during a Black Friday event, which resembles an emergency evacuation (November 2017, South Africa). We measure the flow of pedestrians entering the store and found a higher value ($\langle J \rangle =6.7 \pm 0.8\,$p/s) than the usually reported in ``laboratory'' conditions. We performed numerical simulations to recreate this event. The empirical measurements were compared against simulated evacuation curves corresponding to different sets of parameters currently in use in the literature. The results obtained suggest that the set of parameters corresponding to calibrations from laboratory experiments (involving pedestrians in which the safety of the participants is of major concern) or situations where the physical contact is negligible, produce simulations in which the agents evacuate faster than in the empirical scenario. To conclude the paper, we optimize two parameters of the model: the friction coefficient $k_t$ and the body force coefficient $k_n$. The best fit we found could replicate the qualitative and quantitative behavior of the empirical evacuation curve. We also found that many different combinations in the parameter space can produce similar results in terms of the goodness of fit.