Relational Reinforcement Learning Based Autonomous Cell Activation in Cloud-RANs

Abstract

The emergence of future 5G technologies has given cloud radio access networks (C-RANs) considerable attention. In the C-RANs, distributed remote radio heads (RRHs) are connected to centralized baseband units (BBUs) which have high capacity processors through radio links to forward radio signals from users. For the BBU pool to control energy consumption and user satisfaction levels, reinforcement learning techniques become the best option. In this paper, we propose an autonomous cell activation framework and customized physical resource allocation schemes to balance energy consumption and QoS satisfaction in wireless networks. We formulate the cell activation problem as a Markov decision process and set up a relational reinforcement learning model based on online k-means clustering and anchor-graph hashing (AGH) to satisfy the user QoS demand and to achieve low energy consumption with the minimum number of the active RRHs under varying traffic demand and user mobility. The extensive simulations are conducted to show the effectiveness of our proposed solution under a mobility scenario compared with the state-of-the-art schemes.

Keywords

• Relational reinforcement learning
• anchor graph hashing
• online k-means clustering
• autonomous cell activation
• cloud radio access networks