Placing Agents in Massively Multi-Agent Systems

In this paper, we propose a technique which determines the placement of agents, taking both the computation load of agents and the interaction among agents in distributed multi-agent systems into account. We formulate the agent placement problem as the problem that the load of agents is distributed among multiple agent servers and the interaction cost arising among agent servers is minimized. Then, we provide an approximate algorithm to the agent placement problem by means of using an idea of an approximate algorithm for network partition problem. In this paper, we conduct a computer simulation to test the effect of our technique.