Research On Agent-based Simulation Platform For Crowd Evacuation Considering Building Fire Impacts

The research on crowd evacuation under building fire scenarios is of great significance for ensuring people's safety and reducing casualties.With the development of computing technologies,the modeling and simulation for crowd evacuation has now become the main method for evacuation research.However most of the evacuation models ignore the effects of fire and its products on evacuees,which inevitably leads to inaccuracy in simulation results.Although several models exist that consider the effects of fire on evacuees,the evacuees' behaviors modeled by them are highly simplified.Multi-agent system(MAS)is a typical kind of evacuation model.Expressing macro emergent characteristics of the whole system through the modeling of individual micro-behaviors,MAS is suitable for simulating the evacuees' behaviors.Therefore,the purpose of this study is to use multi-agent modeling to establish an evacuation simulation model that considers the real-time impact of building fire,in order to achieve a more accurate and realistic simulation of crowd evacuation under building fire scenarios.This study comprehensively summarized the impact mechanisms of fire and its products on the evacuees' physiological conditions and behavioral decisionmaking through literature review,and transformed them into behavior rules implementable by multi-agent modeling.For physiological aspects,the impacts of fire hazards on visibility,speed of movement,and health level of evacuees are considered.For behavioral decision-making,the evacuation mechanism for avoiding danger is mainly considered.To implement the above model,secondary development of an existing MAS-based evacuation platform called SAFEgress was carried out,and a new simulation platform,Fire Risk Emulated Environment for Egress(FREEgress),was developed.The tool is able to interact with fire simulation software Pyrosim,and simulate crowd evacuation by considering the real-time impact of fire.The simulation results for evacuation can be visualized synchronously with the fire spread process.This study validated the effectiveness of FREEgress by comparing its simulation results with SAFEgress and FDS+Evac,which is a widely used commercial simulation software for fire evacuation.After the validation,a case study was conducted using FREEgress.During the case study,the impacts of fire locations,delay time and human behavior patterns on the evacuation process were studied by conducting largescale evacuation simulation experiments.The simulation results show that the above three factors each have impact on evacuation,while the existence of the other two factors may change the influence of single factor on evacuation.Based on the simulation results,this study proposes specific suggestions for fire emergency management.To sum up,this study modeled the mechanism of fire impact on evacuees,and then developed an evacuation simulation platform that considers the realtime impact of fire,leading to more accurate and realistic simulation of crowd evacuation under building fire scenarios.The results of this study will be of significant contribution to the understanding of evacuation behavioral patterns of the crowds under fire emergency situation.Practically the results will also be of significant value to emergency management professionals in predicting and managing building fire emergency evacuation processes.Moreover,the results can be used to facilitate the evaluation of evacuation performance and design optimization of performance-based fire protection for certain buildings.