Reinforcement Learning-Based Joint Cooperation Clustering and Content Caching in Cell-Free Massive MIMO Networks

Abstract

This paper studies the previously unexamined problem of joint cooperation clustering and content caching in a cache-enabled cell-free massive multiple-input multiple-output (CF-mMIMO) network that comprises a large number of access points (APs) collaboratively serving users without cell structure limitations. A joint cooperation clustering and content caching design is motivated by the observation that forming cooperation clusters (i.e., determining the sets of serving access points (APs) for users) based on channel quality alone or caching status alone is suboptimal. We develop a deep reinforcement learning (DRL)-based joint design scheme for dynamic CF-mMIMO networks. The proposed scheme demonstrates favorable network energy efficiency (EE) performance and does not require prior information such as user content preferences.