QMIX: Monotonic Value Function Factorisation for Deep Multi-Agent Reinforcement Learning

TL;DR

QMIX introduces a novel value-based multi-agent reinforcement learning method that enables decentralized policies to be trained centrally with guaranteed consistency, significantly improving performance on complex tasks.

Contribution

QMIX proposes a monotonic value function factorization approach that allows scalable, centralized training of decentralized policies with guaranteed consistency.

Findings

QMIX outperforms existing methods on StarCraft II tasks.

The monotonicity constraint enables efficient joint action-value maximization.

QMIX achieves better coordination among agents in complex environments.

Abstract

In many real-world settings, a team of agents must coordinate their behaviour while acting in a decentralised way. At the same time, it is often possible to train the agents in a centralised fashion in a simulated or laboratory setting, where global state information is available and communication constraints are lifted. Learning joint action-values conditioned on extra state information is an attractive way to exploit centralised learning, but the best strategy for then extracting decentralised policies is unclear. Our solution is QMIX, a novel value-based method that can train decentralised policies in a centralised end-to-end fashion. QMIX employs a network that estimates joint action-values as a complex non-linear combination of per-agent values that condition only on local observations. We structurally enforce that the joint-action value is monotonic in the per-agent values, which allows tractable maximisation of the joint action-value in off-policy learning, and guarantees consistency between the centralised and decentralised policies. We evaluate QMIX on a challenging set of StarCraft II micromanagement tasks, and show that QMIX significantly outperforms existing value-based multi-agent reinforcement learning methods.