Abstract
This paper addresses an advanced cooperative theory for Multi-Agent Systems based on Sheepdog simulation. A purpose of this simulation is for cooperative swarm handling. Most of conventional studies of MAS focused on one target handling by multiple agents. If the target were multiple agents, the cooperation would be complex. However, this cooperative theory would be more available for practical applications such as robot motion planning for prevention of forest fire spreading. As first of the challenge, simulation platform and evaluation technique are described.
