Energy-efficient communications through self-supervised learning and distributed wireless networks

Given the ever-growing demand for high throughput wireless communications, the power consumption used by these networks continues to soar. Many research efforts have been focused on providing higher throughput, reliability and lower latencies. However, energy consumption remains an afterthought rather than a first principal goal in the development of wireless technologies. Given the growing concerns regarding carbon emissions and sustainability, optimizing for energy-efficiency should be put forward as a core component of developing communication technologies. As such, this research focuses on reducing the dynamic energy consumption of (distributed) massive multiple-input multiple-output (MIMO) systems. One major bottleneck for the energy-efficiency is the low efficiency of the power amplifiers (PA). In order to overcome this hurdle, novel machine learning based precoders and decoders will be developed. The main goal of these precoders/decoders is to allow the PA to work at a more energy-efficient operating point closer to saturation by mitigating the negative effects of the increased nonlinear distortion. Additionally, new machine learning based channel estimation methods will be developed which can either directly or indirectly reduce the energy consumption. An example of this could be to avoid the explicit channel estimation process and directly take the pilots as inputs to a neural network precoder. This avoids additional computations which reduces the energy consumption.