Evacuation simulation supporting high level behaviour-based agents

Abstract

Computer based models describing pedestrian behaviour in an emergency evacuation play a vital role in the development of active strategies that minimise the evacuation time when a closed area, with a relatively small number of fixed exits, must be evacuated for a large number of people. The proposed model has a hybrid structure where the dynamics of fire and smoke propagation are modelled by mean of Cellular Automata and for simulating peoples behaviour we use Intelligent Agents. Each agent will possess certain psychological, physiological and social characteristics and based on information that is capable of receiving from its sensors, it may perceive what is happening around, and then take a decision that will reflect its ability to cope with the emergency evacuation, called in this work, behaviour. The simulation model consists of two sub-models, called pedestrian and environmental. As part of the pedestrian model, we have prototyped a methodology that is able to model some of the frequently observed human behaviours in evacuation exercises. In order to test the developed behaviours, we choose simple exercises where the model is applied to slightly different situations of an evacuation due to a potential hazard, such as fire, smoke or some kind of collapse.