A review of machine learning techniques for enhanced energy efficient 5G and 6G communications

Abstract

Cellular technologies have evolved continuously from the 1st to the 5th generation (5G) to meet the exponentially growing needs of bandwidth, throughput and latency. However, the energy consumption experienced a proportional rise generation-wise when new hardware to support additional applications were incorporated. 5G, which already consumes four times more energy than 4G, is expected to bring about a significant spike in the conventional trajectory of energy consumption. 5G is therefore triggering a major concern for energy efficiency and with an even higher technical and network complexity, 6G will pause an unprecedented challenge to energy efficiency and sustainability. This paper focuses on the energy consumption at the base station and access network levels, which amount to around 80% of energy consumption in mobile networks. Machine learning techniques can be employed in several ways to improve the energy efficiency in these components. In this paper, efficient base station deployment strategies and adaptive operational modes as well as access network technologies such as massive MIMO and millimeter waves, which employ machine learning to enhance the energy efficiency, have been reviewed in depth. Since existing research works have focused mostly on a single optimization strategy at either the base station or access network level, this paper proposes a framework, which combines efficient base station deployment methods and machine learning-based switching between different operation modes based on traffic load. Moreover, an adaptive beamforming methodology through identification of hotspots and user association, sub-channel and power allocation in heterogeneous networks is discussed in detail.