Scalability in distributed simulations of agent-based models

Research on systems of autonomous agents, called multiagent systems (MAS), has received much interest in the domain of (distributed) artificial intelligence in recent years. MAS are most suitable for the development of distributed applications within an uncertain and dynamically changing environment (Logan 2005). For validation of such systems agent based simulation is a new modeling paradigm not limited to systems which qualify as MAS by default. The focus of the work presented here is on scalability aspects of simulation environments for agent based simulations. Scalable solutions are required, as complex models require the capability to simulate hundreds or more complex deliberative agents. This is a capability which is often lacking in existing simulation environments for agents. We investigate different aspects which influence scalability and present a solution for enabling a scalable and efficient distributed simulation of agent-based models based on an adapted optimistic synchronization protocol which limits the level of optimism by using knowledge about agent interaction patterns.