User-Centric Clustering in Cell-Free MIMO Networks using Deep Reinforcement Learning

The canonical setup of cell-free massive multiple-input multiple-output (MIMO), where all the access points (APs) serve all the users, does not scale well. In this work, we propose a deep reinforcement learning (DRL) approach to user-centric clustering in which each user is served by only a subset of APs. The clusters are formed such that either a given user demand is satisfied or the network sum rate is maximized. Unlike previous studies, we allow the clusters to vary in size depending on the propagation conditions. We design our DRL framework to be flexible enough to accommodate different performance targets in terms of the sum spectral efficiency, fronthaul capacity and power consumption. By optimizing the AP selection for each user, our proposed scheme is able to achieve the same performance as the canonical setup (upper bound) with significantly lower fronthaul requirements.