QTRAN: Learning to Factorize with Transformation for Cooperative Multi-Agent Reinforcement Learning
This addresses a limitation in MARL for cooperative tasks by enabling more general factorization, though it is incremental as it builds on existing methods like VDN and QMIX.
The paper tackles the problem of factorizing joint action-value functions in multi-agent reinforcement learning (MARL) under centralized training with decentralized execution, proposing QTRAN, which removes structural constraints like additivity and monotonicity to cover a wider class of tasks and demonstrates superior performance in experiments, with larger margins in games that penalize non-cooperative behavior.
We explore value-based solutions for multi-agent reinforcement learning (MARL) tasks in the centralized training with decentralized execution (CTDE) regime popularized recently. However, VDN and QMIX are representative examples that use the idea of factorization of the joint action-value function into individual ones for decentralized execution. VDN and QMIX address only a fraction of factorizable MARL tasks due to their structural constraint in factorization such as additivity and monotonicity. In this paper, we propose a new factorization method for MARL, QTRAN, which is free from such structural constraints and takes on a new approach to transforming the original joint action-value function into an easily factorizable one, with the same optimal actions. QTRAN guarantees more general factorization than VDN or QMIX, thus covering a much wider class of MARL tasks than does previous methods. Our experiments for the tasks of multi-domain Gaussian-squeeze and modified predator-prey demonstrate QTRAN's superior performance with especially larger margins in games whose payoffs penalize non-cooperative behavior more aggressively.