System issues in multi-agent simulation of large crowds

Crowd simulation is a complex and challenging domain. Crowds demonstrate many complex behaviours and are consequently difficult to model for realistic simulation systems. Analyzing crowd dynamics has been an active area of research and efforts have been made to develop models to explain crowd behaviour. In this paper we describe an agent based simulation of crowds, based on a continuous field force model. Our simulation can handle movement of crowds over complex terrains and we have been able to simulate scenarios like clogging of exits during emergency evacuation situations. The focus of this paper, however, is on the scalability issues for such a multi-agent based crowd simulation system. We believe that scalability is an important criterion for rescue simulation systems. To realistically model a disaster scenario for a large city, the system should ideally scale up to accommodate hundreds of thousands of agents. We discuss the attempts made so far to meet this challenge, and try to identify the architectural and system constraints that limit scalability. Thereafter we propose a novel technique which could be used to richly simulate huge crowds.