Understanding and Simulating Large Crowds

Simulation tools are often used to establish pedestrian and evacuee performance. The accuracy and reliability of such tools are dependent upon their ability to qualitatively and quantitatively capture the outcome of this performance. This paper investigates the relationship between the representation of low-level agent actions and the generation of reliable emergent, high-level conditions that can then be used to better understand the conditions that may develop in large crowds and mitigate against them. Data has been collected concerning the movement of pilgrims during the Hajj. This paper presents a simple framework for categorizing these real-world observations and then translating them into the simulated environment – extracting key information from the data collected to configure the simulation tool as required. Several scenarios are simulated using the buildingEXODUS model to test the impact of representing these observations to a greater or lesser degree. This enables the importance of low-level behaviours upon emergent conditions to be investigated, even when simulating large numbers of pilgrims attending the Hajj; i.e. in large crowds. The relationship between low-level agent actions and the high-level emergent conditions is investigated using analytical and simulation tools. This paper should help future researchers (1) identify and extract key factors from crowd data and then (2) appropriately configure simulation tools to represent agent behaviour and the subsequent emergent conditions produced (e.g. congestion, flow patterns, etc.).