Joint Cooperation Clustering and Downlink Power Control for Cell-Free Massive MIMO with Deep Reinforcement Learning

Recently, numerous clustering and power control approaches have been developed to optimize network performance for cell-free massive multiple-input multiple-output (MIMO) systems. Notably, deep reinforcement learning (DRL) has emerged as a promising solution for network optimization. This paper investigates the joint cooperation clustering and downlink power control problem in cell-free massive MIMO, utilizing the deep deterministic policy gradient (DDPG) algorithm to maximize the long-term global average proportional fairness (PF). Our results demonstrate that the proposed DDPG method effectively outperforms the baselines in both clustering and power control. Furthermore, the proposed scheme exhibits robustness as it maintains superior performance even as the number of clusters varies.