Joint AP Scheduling and Power Allocation Based on Synergistic DRL for Cell-Free Massive MIMO

Abstract

The challenges of energy consumption posed by 5G have emerged as a critical bottleneck for next generation mobile communications. In response to the “Dual Carbon” initiative, we focus on enhancing downlink energy efficiency (EE) in cell-free massive MIMO systems. Unlike most existing studies, which overlook the dynamic fluctuations in users’ downlink rate demands, we aim to optimize the overall downlink energy efficiency while maintaining a constrained satisfaction ratio for users’ spectral efficiency (SE) requirements. In this letter, we propose a synergistic Deep Reinforcement Learning (DRL) cell-free framework, which utilizes the Advantage Actor-Critic (A2C) to jointly and dynamically adjust the idle/active states of access points (APs) and allocate the transmitting power. Simulation results demonstrate that the synergistic A2C-based scheme with idle/active scheduling can effectively improve the energy efficiency of cell-free massive MIMO system, while ensuring the satisfaction of spectral efficiency requirements.