[4G3-GS-7] Agents: Multi-agent system (2)

[4G3-GS-7-02] Team Formation Based on Reciprocity in Unbalanced Workload Environments

〇Koki Sato¹, Ryoya Funato¹, Toshiharu Sugawara¹ (1. Waseda University)

Keywords:multi-agent system, task allocation, team formation

This paper proposes a method for task allocation in the framework of the team formation problem in a non-uniform environment in which agents are scattered in a two dimensional metric space and are required some delays for communications proportional to the distance.
We already proposed a task-oriented team formation method by generating a cooperative structure based on the learning of the reciprocity among agents and that of the roles in individual teams to improve the success rate of forming teams. However, our previous method assumed that tasks occur in an environment uniformly, and so, it is unknown, when tasks occur in an unbalanced manner, if agents can form effective teams by taking into account their communication delays and agents' performance/capabilities of task execution. Therefore, we introduced a model in which the environment consists of a number of areas that have different workload, and then, extended our previous team formation method so that manager agents were able to invite distant but appropriate agents with high capabilities to join as the reciprocal team members. We experimentally evaluated our method and investigated how managers selected their team members in the unbalanced dynamic environments.