Local Interaction of Agents for Division of Labor in Multi-agent Systems

Task allocation problem has been an issue in multi-agent systems. Among many interesting tasks, we focus on an algorithm for the proportional regulation of population where the swarm is divided into groups depending on task demands. We take the response threshold model inspired by division of labor in several social insects. In our approach, the member proportion of each sub-group is regulated proportional to the external task demands and local social interactions among agents. Here, the interactions control the response thresholds for given tasks. The proposed algorithm was applied to simulation experiments of robots, and the experimental results show that the proposed method has adaptive and robust responses under dynamically changing environments.