Radio Resource Management for New Application Scenarios in 5G: Optimization and Deep Learning

Abstract

The fifth-generation (5G) New Radio (NR) systems are expected to support a wide range of emerging applications with diverse Quality-of-Service (QoS) requirements. New application scenarios in 5G NR include enhanced mobile broadband (eMBB), massive machine-type communication (mMTC), and ultra-reliable low-latency communications (URLLC). New wireless architectures, such as full-dimension (FD) massive multiple-input multiple-output (MIMO) and mobile edge computing (MEC) system, and new coding scheme, such as short block-length channel coding, are envisioned as enablers of QoS requirements for 5G NR applications. Resource management in these new wireless architectures is crucial in guaranteeing the QoS requirements of 5G NR systems. The traditional optimization problems, such as subcarriers and user association, are usually non-convex or Non-deterministic Polynomial-time (NP)-hard. It is time-consuming and computing-expensive to find the optimal solution, especially in a large-scale network. To solve these problems, one approach is to design a low-complexity algorithm with near optimal performance. In some cases, the low complexity algorithms are hard to obtain, deep learning can be used as an accurate approximator that maps environment parameters, such as the channel state information and traffic state, to the optimal solutions. In this thesis, we design low-complexity optimization algorithms, and deep learning frameworks in different architectures of 5G NR to resolve optimization problems subject to QoS requirements. First, we propose a low-complexity algorithm for a joint cooperative beamforming and user association problem for eMBB in 5G NR to maximize the network capacity. Next, we propose a deep learning (DL) framework to optimize user association, resource allocation, and offloading probabilities for delay-tolerant services and URLLC in 5G NR. Finally, we address the issue of time-varying traffic and network conditions on resource management in 5G NR.