摘要

This paper offers an analysis of cluster formations on planer cells comprised of multi-agents utilizing local interactions and state transitions based on Genetic Programming (GP) and its applications. First, we illustrate that if the states of agents are allowed to have continuous values, equilibrium is attained on the basis of the fixed-point theorem. We also show that if the agents are restricted to binary states, equilibrium is attained in an asymptotic sense. However, for agents characterized by more than one state, the attainment of equilibrium is not ensured. We examine our results by using a simulation wherein agents learn from past experiences based on GP. Finally, we demonstrate a system comprised of cluster formations on planer cells comprised of artificial agents, and apply this system to the clustering of employees in firms.