Facilitating Cellular-Automata-Based Computational Space for Parallel Multi-Agent Simulations

The multi-agent simulation is viewed as the interaction of agents, and is frequently applied for analyzing their selforganization in complex systems. To decide their next behavior, such agents generally access spatial information that needs to be not only made available to the agents in local but also diffused over an entire simulation space. As an efficient implementation of information diffusion, we propose to combine the principle of cellular automata into the multi-agent simulation. There, agents walk over cellular automata that diffuse spatial information. To evaluate its efficiency, we have facilitated such information diffusion on top of the M++ parallel multi-agent simulator. Our performance experiment showed that implementing cellular automata in the Codi-1bit multi-agent simulation achieved 2.18time speed-up when using 8 processors, as compared to the one with no support of cellular automata.