Agent-based simulation of crowds in indoor scenarios

Abstract

Crowd simulation models usually focus on performance issues related with the management of very large numbers of agents. This work presents an agent-based architecture where both performance and flexibility in the behaviour of the entities are sought. Some algorithms are applied for the management of the crowd of agents in order to cope with the performance in the processing of their movements and their representation, but at the same time some alternative reasoning mechanisms are provided in order to allow rich behaviours. This facilitates the specification of different types of agents, which represent the people, sensors and actuators. This is illustrated with a case study of the evacuation of the building of the Faculty of Computer Science, where different types of human behaviours are modelled for these situations. The result is the simulation of more realistic scenarios.