Few-Shot Teamwork

TL;DR

The paper introduces the few-shot teamwork (FST) problem, where pre-trained agents from different tasks must quickly adapt and collaborate on unseen tasks, advancing multi-agent learning and ad hoc teamwork.

Contribution

It defines the FST problem, highlighting its significance and potential to improve rapid adaptation and collaboration in multi-agent systems.

Findings

FST formalizes the challenge of quick adaptation in multi-agent teams.

Identifies two core problems: reducing training experience and enabling collaboration with unfamiliar agents.

Suggests FST as a pathway to progress in multi-agent reinforcement learning and ad hoc teamwork.

Abstract

We propose the novel few-shot teamwork (FST) problem, where skilled agents trained in a team to complete one task are combined with skilled agents from different tasks, and together must learn to adapt to an unseen but related task. We discuss how the FST problem can be seen as addressing two separate problems: one of reducing the experience required to train a team of agents to complete a complex task; and one of collaborating with unfamiliar teammates to complete a new task. Progress towards solving FST could lead to progress in both multi-agent reinforcement learning and ad hoc teamwork.