Analyzing large-scale crowd simulations for building evacuation

Abstract

Animated virtual crowds have been used last years for analyzing human factors in scenarios where masses of people gather. A typical example is building evacuation in case of fire. Scalability still remains as an open issue for these multi-agent systems applications. In this paper, we use a scalable architecture to simulate a large-scale version of a virtual crowd in a building evacuation. From the social point of view, the results provided by the large-scale version of the crowd add new and crucial information about the agents behavior, emphasizing the need for a small amount of trained leaders in order to save lives. From the system point of view, the results show that the trend of avatars towards crowding in some areas highly increases the computation time for the agents hosted in some client computers. Therefore, this trend should be taken into account when designing large-scale evacuation simulations.