System Issues in Crowd Simulation using Massively Multi-Agent Systems

Crowd Simulation has emerged as an interesting and challenging research domain in recent years. A number of multi-agent based simulation systems have been proposed notably RoboRescue. In this paper we focus on the scalability issues for such a rescue simulation system. Simulating large crowds in rescue systems throws up many challenges. Crowd events and their associated phenomenon are difficult to model. Different types of crowd simulation systems have been developed, ranging from force-modelling approaches, cellular automata based simulations and rule-based architectures. However, multi-agent paradigm is particularly suitable for modelling human behaviour, as human characteristics can be objectively mapped to agent behaviour. In this paper we analyze the scalability issues for multi-agent based crowd simulation systems. Scalability of a such a system is vital as the system is required to simulate very large-sized crowds. To realistically model a disaster scenario for a large city the system should ideally scale up to accommodate hundreds of thousands of agents.