An agent-based model for simulating cooperative behavior in crowd evacuation during toxic gas terrorist attacks

Abstract

Toxic gas leaks pose severe threats to public safety and societal stability, leading to large-scale casualties and social panic. This paper focuses on crowd evacuation behavior during toxic gas leak incidents, proposing an evacuation model that combines Computational Fluid Dynamics (CFD) and Agent-Based Modeling (ABM). By introducing a helping mechanism among agents with prosocial personalities, the study examines the impact of the prosocial personality ratio (p) on evacuation time, fatalities, and severe injuries. Subsequently, the effects of the p under varying conditions, such as total population size and evacuation response time, are explored. Additionally, a Random Forest model is employed to accurately predict evacuation risks, and the NSGA-III multi-objective optimization algorithm is utilized to identify the optimal range of p across different scenarios. The results indicate that a reasonable proportion of prosocial personalities can significantly reduce fatality rates and enhance overall evacuation efficiency. However, an excessively high proportion of prosocial individuals may increase crowd casualties due to extended delays caused by helping behaviors. This study contributes to the body of knowledge on public safety, provides methodological references for developing evacuation strategies during toxic gas diffusion incidents, and offers guidance for future emergency management practices.