Downlink Power Control in Dense 5G Radio Access Networks Through Deep Reinforcement Learning

During the past decades, a myriad of inter/intracell interference mitigation techniques has been suggested for different wireless technologies. Nevertheless, the concept of downlink power control for interference mitigation has yet to be explored in 5G radio access networks. In this paper, we propose a data-driven approach based on deep reinforcement learning for downlink power control in dense 5G networks. The solution builds upon the well-known DQN algorithm and its recent extensions, aiming to maximize user rates. Using a 5Gcompliant system-level simulator, we compare the performance of our proposed method to fixed power allocation approaches. Test results show that the proposed method is successful at improving data rates at the cell-edge while reducing total transmitted power compared to the baseline.