SEEDRL: Smart Energy Efficiency Using Deep Reinforcement Learning for 6G Networks

The rapid increase in mobile network traffic has led to the dense deployment of network cells and the introduction of technologies such as massive Multiple-Input Multiple-Output (m-MIMO) to achieve high gain and spectral efficiency. However, these benefits come with a significant growth in Operational Expenditure (OPEX) and energy consumption, which remains a major challenge in beyond 5G and 6G networks. In this paper, we employ Deep Reinforcement Learning (DRL) techniques to efficiently switch off cells and mute MIMO antenna elements at some specific times to achieve a higher gain in terms of Energy Saving (ES) and the number of network changes without majorly affecting user satisfaction. Through extensive experiments, we demonstrate that our proposed method called Smart Energy Efficiency using DRL (SEEDRL) saves power by 8.99% and significantly reduces the number of ES state changes by 22.83% compared to its counterpart the threshold-based algorithm.