Energy-Efficient Access Point Sleep Control in User-Centric Cell-Free Massive MIMO Systems

In this paper, the problem of energy efficient access point (AP) sleep control in user-centric cell-free massive MIMO (CFmMIMO) systems is investigated. The large-scale deployment of AP in CFmMIMO makes it challenging to obtain an effective AP sleep policy in real time. For this consideration, we propose to use a proximal policy optimization (PPO) based deep reinforcement learning method, which has advantages in solving high-dimensional problems, to obtain an energy-efficient AP sleep strategy. We map the continuous action generated by PPO agent into discrete domain and design a specific reward for PPO to train which guarantees the quality of service (QoS) requirement of users. Simulation results show that our proposed method outperforms the state-of-art benchmark methods.