Enhanced Non-Orthogonal Multiple Access (NOMA) for 5G Wireless Systems based on Deep Learning

点击翻译按钮获取中文摘要

In the proposed research, we will develop new and effective mobile communication techniques by utilizing deep learning technology. The wireless industry is currently facing many challenges in developing fifth generation (5G) wireless systems: extremely high data rates, extremely short transmission delay, and exponentially increasing mobile data traffic. For example, in 5G, network-level data rate should be 10-20 Gbps (10-20 times higher compared to the current 4G) and the latency (the end-to-end transmission delay) should be one millisecond (one-fifth compared to the current 4G). Furthermore, it is expected that Canadian mobile data traffic will grow 500% from 2016 to 2022 at a compound annual growth rate of 38%. In order to address the challenges, many researchers have studied a fundamentally different way of communications in which multiple users are simultaneously allowed to transmit (or receive) their data by non-orthogonally using the system resources (e.g., time, frequency, codes). This technique is referred to as the non-orthogonal multiple access (NOMA), which can theoretically provide much higher data rate and shorter delay. However, the NOMA system design and optimization is very challenging, because the system is highly nonlinear and very complicated. In order to overcome the critical issues of the NOMA systems, this research proposes to take a new and intelligent approach by exploiting state-of-the-art deep learning techniques. Deep learning is a branch of machine learning that has the capability of learning useful representation for data by extracting complicated distributed features automatically, resulting in very powerful nonlinear systems. Deep learning is an important component of artificial intelligence (AI), which has been shown to help advance a wide variety of research areas. By utilizing state-of-the-art deep learning techniques, we will develop NOMA systems that can autonomously operate and support very high data rates. Specifically, using deep learning, we aim to resolve four major challenges of NOMA: error mitigation, user grouping, precoding, and adaptive control. According to a new report from Accenture, the new 5G wireless system will contribute $40 billion annually to Canada's economy by 2026. Furthermore, in Canada's wireless industry, more than 150,000 short-term jobs will be created from 2020 to 2025, and an additional 250,000 permanent jobs will be annually created by 2026. The proposed research will enable Canadian wireless companies to grow fast and to lead the global wireless industry. Most of all, the proposed research is “deliberately” and “strategically” designed to train HQP in two key areas at the same time: i) wireless communications and ii) deep learning. In the proposed research program, 4 MSc students and 4 PhD students will be trained to learn the state-of-the-art for future wireless systems and the fundamentals of deep learning.