Research Issues in Agent-Based Simulation for Pedestrians Evacuation

Abstract

Crowd evacuation in emergencies may lead to fatalities if the evacuation plans were not tested and evaluated. Traditionally, evacuation drills have been, and still are, being used to assess evacuation plans. However, in recent years the simulation of evacuation plans during emergencies has emerged as a strong alternative that is cost effective and potentially more accurate. Agent-Based Simulation (ABS) is the preferred type of simulation for evacuation scenarios, due to its ability to model individual decision-making and social behaviour. In this paper we conduct meta-analysis of eighty-one peer-reviewed papers published between 2009 and 2019 that used ABS to model pedestrian evacuation. Our analysis assesses the current state-of-art and identifies opportunities for improvement. We identify seven dimensions over which the surveyed papers agree or differ. The dimensions include purpose of the simulation, type of emergency and environment considerations, type and scale of evacuated space, simulation software used, agents' characteristics and behaviour, support of evacuation policies, and analysis and validation. We conduct meta-analysis of the surveyed papers along the identified dimensions. One of the main findings of our analysis is the lack of a standardized validation methodology for ABS of emergency evacuation.