Transferable Multi-Agent Reinforcement Learning with Dynamic Participating Agents

TL;DR

This paper introduces a novel multi-agent reinforcement learning framework with a network architecture and few-shot learning algorithm that enables rapid adaptation to changing agent participation during training, significantly reducing retraining time.

Contribution

The paper proposes a new network architecture and few-shot learning algorithm for MARL that allows agents to join or leave during training without retraining from scratch.

Findings

Model adaptation when new agents join is over 100 times faster.

The approach is applicable to cooperative, competitive, and mixed settings.

The method maintains training effectiveness despite dynamic agent participation.

Abstract

We study multi-agent reinforcement learning (MARL) with centralized training and decentralized execution. During the training, new agents may join, and existing agents may unexpectedly leave the training. In such situations, a standard deep MARL model must be trained again from scratch, which is very time-consuming. To tackle this problem, we propose a special network architecture with a few-shot learning algorithm that allows the number of agents to vary during centralized training. In particular, when a new agent joins the centralized training, our few-shot learning algorithm trains its policy network and value network using a small number of samples; when an agent leaves the training, the training process of the remaining agents is not affected. Our experiments show that using the proposed network architecture and algorithm, model adaptation when new agents join can be 100+ times faster than the baseline. Our work is applicable to any setting, including cooperative, competitive, and mixed.