Energy Efficiency in 5G Communications – Conventional to Machine Learning Approaches

Paper Energy Efficiency in 5G Communications – Conventional to Machine Learning Approaches Muhammad Khalil Shahid, Filmon Debretsion, Aman Eyob, Irfan Ahmed, and Tarig Faisal Higher Colleges of Technology, Abu Dhabi, United Arab Emirates https://doi.org/10.26636/jtit.2020.146820 Abstract—Demand for wireless and mobile data is increasing along with development of virtual reality (VR), augmented reality (AR), mixed reality (MR), and extended reality (ER) applications. In order to handle ultra-high data exchange rates while offering low latency levels, fifth generation (5G) networks have been proposed. Energy efficiency is one of the key objectives of 5G networks. The notion is defined as the ratio of throughput and total power consumption, and is measured using the number of transmission bits per Joule. In this paper, we review state-of-the-art techniques ensuring good energy efficiency in 5G wireless networks. We cover the base-station on/off technique, simultaneous wireless information and power transfer, small cells, coexistence of long term evolution (LTE) and 5G, signal processing algorithms, and the latest machine learning techniques. Finally, a comparison of a few recent research papers focusing on energy-efficient hybrid beamforming designs in massive multiple-input multiple-output (MIMO) systems is presented. Results show that machine learningbased designs may replace best performing conventional techniques thanks to a reduced complexity machine learning encoder. Keywords—5G, energy efficiency, wireless networks. 1. Introduction Conventional fuels used for power generation, heating and transport have contributed to an 80% increase in greenhouse gas emissions compared to 1970. According to projections pertaining to 2040, global energy demand is expected to increase even further, by 30%, with the pace of growth being even faster in developing countries [1]. 5G networks would inevitably be responsible for an increase in the amount of energy used by consumers, therefore contributing to climate change. As the amount of space within the wave spectrum in which consumer devices may operate is increased by the use of millimeter waves, energy usage grows as well, leading to faster global warming. 3GPP standards, including those related to 5G networks, aim to increase capacity and coverage of the system, with energy efficiency gains considered at architectural and functional level [2]. Ensuring that hardware is capable of working within extended operating condition ranges (temperature and humidity levels prevailing in rooms in which equip- ment is located) may lead to a decrease in the amounts of power consumed by air conditioning systems. Small cells used to provide 5G connectivity are claimed to be energy efficient and powered in a sustainable way. However, maintenance- and production-related issues may cause considerable cost implications [1]. Deployment of 5G systems is also expected to improve energy efficiency (EE) of the entire industry as a whole, as the cost of energy per bit of data transferred is, in 5G, equal to one tenth of the level experienced in 4G [2]. However, base stations still remain energy-hungry locations of the network, due to the foreseeable increase in traffic that is expected to grow by several thousand percent. A few papers exist that focus on analyzing EE of 5G networks. Report [3] surveys various optimization techniques, the game theory and machine learning approaches that have been proposed for enhancing power allocation to downlink and uplink channels. Other energy-saving approaches are described therein as well. In paper [1], some of the significant examples discussed include deployment of newer radio resource control (RRC) for context signaling and for reducing the number of redundant state changes. Utilization of advanced clustering and caching techniques on the radio access network (RAN) side has been highly valued for improving latency requested by a group of users and for eliminating the factor of clogging the network by a huge number of requests for the same data. Commercial resource sharing between different operators offers encouraging results in terms of reduced deployment costs and good data rates, while ensuring minimum interference. In a paper [4] a detailed discussion of the various advantages and disadvantages of green and energy efficiency techniques is presented, contributing to understanding the ways in which green radio architecture may be used in 5G and future mobile networks, and presenting the challenges that will be encountered in the process. In this paper, three most promising green solutions are analyzed. Extreme mobile broadband (xMBB) is a service characterized by high data rates, low latency communication (LLC), and extreme coverage. Its spectrum resources include lower bands, and new higher bands with large contiguous bandwidth, (license + LSA + LAA). Its target values are defined by the peak data 1 Muhammad Khalil Shahid, Filmon Debretsion, Aman Eyob, Irfan Ahmed, and Tarig Faisal rate of up to 20 Gbps for downlink and 10 Gbps for uplink. Massive machine-type communications (mMTC) is another type of service that offers the following features: dense mobile networks, wide-area coverage and deep penetration. It relies on lower band frequencies and its spectrum resources include licensed shared access (LSA) and licensed-assisted access (LAA). The target connectivity density value equals 1 million devices per square kilometer. The third service offers ultra-reliable and low-latency communication links, such as V2X. Its spectrum resources are based on lower bands, exclusive licenses, and its target user latency quals 0.5 ms. A typical 5G network is shown in Fig. 1. Mathematically, energy efficiency of the base station side is defined as: R EE = [bits/Joule] , (1) η Pt + Pc where, R is the average overall data rate in bits per second, η is the reciprocal of the transmit power (amplifier efficiency), Pt is the transmission power, and Pc is total power dissipated in the transmitter circuit. Fig. 1. Model of a multiuser massive MIMO downlink system. The rest of the paper is organized as follows: in Section 2, on/off techniques are presented and energy harvesting is described. Section 3 presents heterogeneous networks and energy efficiency-related considerations. Energy-efficient physical layer hardware designs are reviewed in Section 4, and machine learning techniques ensuring EE are presented in Section 5. 2. Energy Efficiency Using On/Off Techniques and Energy Harvesting Authors in [5] propose three independent energy efficiency optimization solutions to minimize energy consumption by either forcing idle base stations to go to sleep, or by dynamically adjusting the signal range of base stations through 2 software-defined networking. This means that the stream table of the base stations is reconfigured to modify the connections between users and base stations in order to free as high a number of base stati (...truncated)