Evaluating the Safety of Crowds in Enclosed Spaces by Markovian Agents

Abstract

Different paradigms have been used to model the behaviour of pedestrians in a crowd. While some approaches analyze the modelization of crowd motion from a physical perspective, other techniques prefer to focus on similar abstractions following classic paradigms, such as mobile agents or cellular automata, or even from the point of view of social forces interaction. Particular interest has emerged around the topic of crowd behaviour under condition of stress caused by a dramatic event, such as an earthquake, a terrorist attack or a fire. This paper presents a novel approach including a GSPN to describe the behaviour of an individual, which is subsequently translated into a set of Markovian Agents. The model under study reproduces a scenario where a fire, developed in a closed environment, triggers erratic crowd behaviour. It is possible to perceive that a panic situation can be mitigated by enforcing certain measures involving human leaders or small embedded systems (such as Internet-of-things enabled devices), reducing the level of risk related to safety.