A Scalable Energy Consumption Optimization in Downlink Cell-Free Massive MIMO.

Cell-free massive multiple-input multiple-output (MIMO) eliminates inter-cell interference by cooperatively serving users through distributed access points (APs). Energy efficiency is critical as an individual backhaul link is required to connect each AP to a central processing unit (CPU), and hardware static power supplies and wireless signal transmissions consume huge amounts of energy. AP selection is promising to optimize energy efficiency, where some APs can be deactivated when a large portion of users experience notable propagation loss. However, AP selection alone cannot ensure the minimal rate requirement when the number of users is comparable to or exceeds that of APs. To address this, we propose a scalable energy consumption optimization method by jointly optimizing power allocation, AP selection, and user grouping. We formulate an optimization problem that involves both continuous power allocation variables and discrete AP selection/user grouping variables. We propose a low-complexity method that breaks down the optimization problem into two stages: 1) iteratively refined grouping based on the proposed minimal rate maximization criterion and 2) joint power allocation and AP selection. Numerical results show that the proposed method significantly outperforms several baselines.