Terry Z. Liu

QPLEX: Duplex Dueling Multi-Agent Q-Learning

2020 · arXiv (Cornell University)

We explore value-based multi-agent reinforcement learning (MARL) in the popular paradigm of centralized training with decentralized execution (CTDE). CTDE has an important concept, Individual-Global-Max (IGM) principle, which requires the consistency between joint and local action …