Towards Zero-Shot Coordination between Teams of Agents: The N-XPlay Framework

TL;DR

This paper introduces N-XPlay, a framework for zero-shot coordination in multi-agent systems involving multiple teams, addressing the complexity of real-world scenarios with hierarchical interactions.

Contribution

The paper presents N-XPlay and an N-player Overcooked benchmark to evaluate and improve zero-shot coordination among multiple agents and teams.

Findings

N-XPlay outperforms Self-Play in multi-agent coordination tasks.

Agents trained with N-XPlay better balance intra-team and inter-team coordination.

The N-player Overcooked benchmark enables realistic evaluation of multi-team coordination.

Abstract

Zero-shot coordination (ZSC) -- the ability to collaborate with unfamiliar partners -- is essential to making autonomous agents effective teammates. Existing ZSC methods evaluate coordination capabilities between two agents who have not previously interacted. However, these scenarios do not reflect the complexity of real-world multi-agent systems, where coordination often involves a hierarchy of sub-groups and interactions between teams of agents, known as Multi-Team Systems (MTS). To address this gap, we first introduce N-player Overcooked, an N-agent extension of the popular two-agent ZSC benchmark, enabling evaluation of ZSC in N-agent scenarios. We then propose N-XPlay for ZSC in N-agent, multi-team settings. Comparison against Self-Play across two-, three- and five-player Overcooked scenarios, where agents are split between an ``ego-team'' and a group of unseen collaborators shows that agents trained with N-XPlay are better able to simultaneously balance ``intra-team'' and ``inter-team'' coordination than agents trained with SP.