FACMAC: Factored Multi-Agent Centralised Policy Gradients

TL;DR

FACMAC introduces a factored critic with nonmonotonic capabilities and centralized policy gradients, significantly improving cooperative multi-agent reinforcement learning performance across various complex environments.

Contribution

It presents a novel factored critic with nonmonotonicity and a centralized policy gradient method, enhancing learning capacity and coordination in multi-agent settings.

Findings

FACMAC outperforms MADDPG and baselines on multiple benchmarks.

Nonmonotonic critic factorization enables solving complex tasks.

Centralized policy gradient improves coordination among agents.

Abstract

We propose FACtored Multi-Agent Centralised policy gradients (FACMAC), a new method for cooperative multi-agent reinforcement learning in both discrete and continuous action spaces. Like MADDPG, a popular multi-agent actor-critic method, our approach uses deep deterministic policy gradients to learn policies. However, FACMAC learns a centralised but factored critic, which combines per-agent utilities into the joint action-value function via a non-linear monotonic function, as in QMIX, a popular multi-agent Q-learning algorithm. However, unlike QMIX, there are no inherent constraints on factoring the critic. We thus also employ a nonmonotonic factorisation and empirically demonstrate that its increased representational capacity allows it to solve some tasks that cannot be solved with monolithic, or monotonically factored critics. In addition, FACMAC uses a centralised policy gradient estimator that optimises over the entire joint action space, rather than optimising over each agent's action space separately as in MADDPG. This allows for more coordinated policy changes and fully reaps the benefits of a centralised critic. We evaluate FACMAC on variants of the multi-agent particle environments, a novel multi-agent MuJoCo benchmark, and a challenging set of StarCraft II micromanagement tasks. Empirical results demonstrate FACMAC's superior performance over MADDPG and other baselines on all three domains.