FACMAC: Factored Multi-Agent Centralised Policy Gradients

Peng, Bei ORCID: 0000-0003-0152-3180, Rashid, Tabish, de Witt, Christian A Schroeder, Kamienny, Pierre-Alexandre, Torr, Philip HS and Bohmer, Wendelin
(2021) FACMAC: Factored Multi-Agent Centralised Policy Gradients. In: Thirty-fifth Conference on Neural Information Processing Systems, 2021-12-6 - 2021-12-14, Online.

[img] Text
main_neurips2021.pdf - Author Accepted Manuscript

Download (6MB) | Preview


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.

Item Type: Conference or Workshop Item (Unspecified)
Uncontrolled Keywords: multi-agent reinforcement learning, multi-agent policy gradients
Divisions: Faculty of Science and Engineering > School of Electrical Engineering, Electronics and Computer Science
Depositing User: Symplectic Admin
Date Deposited: 26 Oct 2021 15:51
Last Modified: 05 Apr 2023 19:49
Related URLs:
URI: https://livrepository.liverpool.ac.uk/id/eprint/3141744