Network-Based Message Selection for Efficient Communication of Multi-Agent Collaboration

Seungjun Hyeon, Minho Park, Dong-oh Kang

Abstract: This paper proposes an efficient communication method through network-based message selection to reduce messages in multi-agent collaboration. The proposed method optimizes message communication in the distributed execution phase of multi-agent reinforcement learning through network learning. It addresses the limitations of traditional approaches, which often use a centralized communication controller or employ heuristic methods in distributed applications. The effectiveness of this study is demonstrated by applying it to some multi-agent reinforcement learning algorithms in the StarCraft Multi-Agent Challenge environment.