Towards low-latency and ultra-reliable services in wireless communication

Abstract

Ultra-Reliable and Low-Latency Communications (URLLC) services are critical for a wide range of applications such as industrial automation, intelligent transportation systems, gaming, and Virtual/Augmented Reality (VR/AR). The continuous evolution of wireless technologies, represented by 5G New Radio (NR), Wi-Fi 6 (IEEE 802.11ax), and Wi-Fi 7 (IEEE 802.11be), is driven by the goal of delivering URLLC services under a variety of scenarios. However, this goal comes with inherent challenges, such as optimizing scheduling and resource allocation strategies. This thesis addresses these challenges by introducing intelligent solutions for achieving low latency and ultra-high reliability in both 5G and Wi-Fi technologies. Within the context of 5G NR, this thesis addresses some resource allocation problems - the main contributions detailed in Chapters 3 & 4. Chapter 3 explores multiplexing eMBB and URLLC traffic in 5G downlink transmission, that must satisfy the strict URLLC requirements while maximizing eMBB utility. The resource allocation problem is formulated as utility maximization for the eMBB while attaining proportional fairness and simultaneously satisfying URLLC constraints. This is formulated as an integer programming (IP) problem, offering two approaches for resolution: convex relaxation and a greedy algorithm, to attain superiority over basic round-robin in balancing the utility of eMBB users and meeting URLLC users' latency and reliability requirements. Further, it explores the Integrated Access and Backhaul (IAB) as a method for extending network coverage, offering a cross-layer design for routing and resource allocation in IAB multi-hop networks under the current 3rd Generation Partnership Project (3GPP) 5G standards. We propose a novel entropy-based reinforcement learning integrated with a federated learning mechanism, which improves the overall performance of IAB networks and significantly accelerates convergence speed. Through simulation, the effectiveness of the proposed approach in outperforming baseline algorithms from both latency and reliability perspectives is confirmed. Chapter 4 explores low latency applications in emerging Wi-Fi networks, particularly for gaming and Virtual/Augmented Reality (VR/AR) scenarios. The research addresses two critical aspects: adaptive rate control under fast-changing conditions and optimized resource allocation for the newest Wi-Fi 7 standard. The thesis first presents ADR-X, a novel reinforcement learning based wireless rate adaptation technique. ADR-X capitalizes on the power of online learning to respond effectively to rapidly fluctuating channel conditions, which sudden player movements can trigger during gameplay or variations in background interference. Unlike traditional rate adaptation mechanisms that often incur losses between 5-10%, ADR-X, by predictively choosing the appropriate data rates based on channel measurements, is able to limit packet losses to 10x lower levels. Further, we extend the focus to Wi-Fi 7 (IEEE 802.11be) and investigate the new Multi-Link Operation (MLO) feature introduced in this standard. MLO allows simultaneous transmission on multiple channels, enhancing throughput, lowering latency, and reducing collision probability. This work proposes an optimized cross-layer MLO resource allocation algorithm specifically designed for XR-type burst traffic. Compared to baseline MLO resource allocation based on proportional fairness, the proposed algorithm significantly reduces latency, enabling Wi-Fi networks to serve better applications demanding URLLC services.