Credit Assignment with Meta-Policy Gradient for Multi-Agent Reinforcement Learning

TL;DR

This paper introduces a meta-learning-based mixing network with policy gradient for improved reward decomposition in multi-agent reinforcement learning, demonstrating superior performance on StarCraft II benchmarks.

Contribution

It proposes a novel meta-learning framework for reward decomposition that leverages global information and can enhance existing CTDE methods.

Findings

Outperforms state-of-the-art MARL algorithms on 4 of 5 StarCraft II scenarios.

The method is effective with a simple utility network and further improved with role-based utility networks.

Demonstrates the general applicability of the approach to monotonic mixing networks.

Abstract

Reward decomposition is a critical problem in centralized training with decentralized execution~(CTDE) paradigm for multi-agent reinforcement learning. To take full advantage of global information, which exploits the states from all agents and the related environment for decomposing Q values into individual credits, we propose a general meta-learning-based Mixing Network with Meta Policy Gradient~(MNMPG) framework to distill the global hierarchy for delicate reward decomposition. The excitation signal for learning global hierarchy is deduced from the episode reward difference between before and after "exercise updates" through the utility network. Our method is generally applicable to the CTDE method using a monotonic mixing network. Experiments on the StarCraft II micromanagement benchmark demonstrate that our method just with a simple utility network is able to outperform the current state-of-the-art MARL algorithms on 4 of 5 super hard scenarios. Better performance can be further achieved when combined with a role-based utility network.