Pith. sign in

REVIEW 7 cited by

FACMAC: Factored Multi-Agent Centralised Policy Gradients

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2003.06709 v5 pith:X72JH4KJ submitted 2020-03-14 cs.LG cs.AIstat.ML

FACMAC: Factored Multi-Agent Centralised Policy Gradients

classification cs.LG cs.AIstat.ML
keywords multi-agentfacmaccentralisedpolicyfactoredactioncriticgradients
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
read the original 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.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 7 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Action-Factored Multi-Agent Reinforcement Learning for Scalable Quantum Device Tuning

    cs.LG 2026-07 conditional novelty 6.0

    Online action-space factorization via Kalman-refined cross-capacitance lets shared multi-agent policies zero-shot tune larger quantum-dot arrays with near-constant steps.

  2. Sim2O: Efficient Offline-to-Online MARL via Joint Action Composition

    cs.LG 2026-06 unverdicted novelty 6.0

    Sim2O enables efficient offline-to-online MARL by dynamically blending offline and online action proposals across agents and selecting high-value combinations via a centralized value function without auxiliary objectives.

  3. Robust Instruction Compliance in Cooperative Multi-Agent Reinforcement Learning

    cs.AI 2026-05 unverdicted novelty 6.0

    MAVIC corrects Bellman backups at instruction boundaries by adjusting the incoming objective and restoring continuation value, enabling consistent estimation under stochastic instruction switching in a unified policy.

  4. Robust Instruction Compliance in Cooperative Multi-Agent Reinforcement Learning

    cs.AI 2026-05 unverdicted novelty 6.0

    MAVIC corrects Bellman backups at instruction boundaries by adjusting the incoming objective and restoring continuation value, enabling consistent estimation under stochastic instruction switching in cooperative MARL.

  5. Bridging MARL to SARL: An Order-Independent Multi-Agent Transformer via Latent Consensus

    cs.LG 2026-04 conditional novelty 6.0

    CMAT uses a transformer decoder to produce a high-level consensus vector in latent space, enabling simultaneous order-independent actions by all agents and optimization via single-agent PPO, with superior results on S...

  6. Decoupled Delay Compensation: Enhancing Pre-trained MARL Policies via Learned Dynamics Filtering

    cs.MA 2026-05 unverdicted novelty 5.0

    A decoupled estimator combining gated dynamics learning and recursive Kalman filtering improves robustness of pre-trained MARL policies under stale observations and message loss.

  7. Low Variance Trust Region Optimization with Independent Actors and Sequential Updates in Cooperative Multi-agent Reinforcement Learning

    cs.LG 2026-06 unverdicted novelty 4.0

    Proposes a clipping objective for sequential trust-region updates in independent-actor cooperative MARL that yields a monotonic improvement bound and sub-linear convergence to epsilon-Nash equilibria while reducing ad...