A Multi-Agent DRL Method for Distributed Energy-Efficient Association and Hybrid Precoding in mmWave Cell-Free Massive MIMO Systems

In this letter, we investigate a distributed design integrating user-centric association with local sub-connected hybrid precoding in mmWave cell-free massive MIMO (CF-mMIMO) systems, aiming to maximize the long-term global energy efficiency (EE) under quality-of-service and power budget constraints. The joint optimization problem is formulated as a Markov Game problem and a novel weighted critic update multi-agent twin-delayed deep deterministic policy gradient (WCU-MATD3) algorithm is proposed to solve it, which promotes cooperation among access point agents and reduces power consumption in front-haul links. The results show that the proposed WCU-MATD3 algorithm significantly improves the global EE and the trade-off between spectral efficiency (SE) and EE, providing a practical and stable distributed collaboration framework for dynamic distributed environments.