Protocols and Algorithms for Harmonious Coexistence Over Unlicensed Bands in Next-Generation Wireless Networks

Abstract

The unlicensed spectrum offers tremendous opportunities for mobile network operators (MNOs), whose traffic can be offloaded from licensed bands to unlicensed ones. To realize these opportunities, three new technologies have been proposed: LTE-Unlicensed (LTE-U), 4G LTE Licensed-Assisted Access (LAA), and 5G New Radio Unlicensed (NR-U). Although unlicensed spectrum seems attractive to MNOs, its access is fraught with many challenges that need to be addressed. These include coexistence with legacy unlicensed technologies such as Wi-Fi and Bluetooth, fairness in channel access, and supporting a desired level of quality of service (QoS) in a shared-spectrum environment. Addressing these challenges is critical for achieving harmonious coexistence between various technologies. In this dissertation, we design resource management algorithms and channel access protocols to overcome these challenges. We mainly focus on issues arising in three scenarios: (1) coexistence between Listen-Before-Talk (LBT), e.g., Wi-Fi, and non-LBT systems, e.g., LTE-U, (2) coexistence within LBT systems (e.g., 4G LTE LAA, 5G NR-U, and Wi-Fi), and (3) resource allocation for LBT systems sharing a common network infrastructure (e.g., 5G NR-U and Wi-Fi). For the first scenario, we develop a cross-technology detection scheme that allows LBT devices to concurrently transmit over and sense the channel, a.k.a., full-duplex (FD) sensing. This will reduce the impact of collisions between LBT and non-LBT devices. LBT devices can then detect collisions with non-LBT devices earlier, allowing them to react properly by pausing transmission and preventing interference. We develop a framework based on Partially Observable Markov Decision Process (POMDP) that allows FD-enabled LBT devices to mitigate interference generated by non-LBT devices. By following a POMDP policy, LBT devices can jointly adapt their transmission rates and duplex mode based on their belief about interference caused by non-LBT devices. Although enabling LBT devices with FD capabilities helps mitigate cross-technology interference, it makes the provisioning of QoS harder. Traditional QoS provisioning frameworks are designed to support half-duplex operation. Accordingly, we present a framework, called AFD-QoS, for provisioning of QoS in an FD network. AFD-QoS incorporates FD-based Enhanced Distributed Channel Access (FD-EDCA) and FD-based Block Acknowledgement (FD-BA) schemes to achieve its goals. For the second scenario, we investigate the fair setting of LBT parameters, e.g., contention windows, airtime, etc., and the harmonious setting of their sensing thresholds (STs). To support different QoS applications, e.g., voice, video, etc., LBT systems define multiple channel access classes with different settings. To study the interplay between these classes, we develop a Markov-based model and derive key performance measures, including probability of successful transmission, average channel access delay, and effective throughput for each class. We first discuss how the heterogeneous setting of channel access parameters across traffic classes leads to unfairness in channel access. Our study pinpoints some settings that need to be optimized to ensure fairness. We then discuss the heterogeneity due to having fixed and asymmetric configurations of ST values among different devices, which results in different sensing floors and gives rise to hidden and exposed terminals. To reduce collisions and improve frequency reuse, we investigate distributed learning solutions for adapting the ST values based on the observed environment. We develop a novel clustering-based multi-armed bandit framework, called Sense-Bandits, to perform such adaptation in real time, aiming at boosting the overall network throughput. Finally, we investigate the challenges arising when MNOs run 5G NR-U services over a shared network infrastructure. Ensuring fairness and efficient allocation of network resources among MNOs are challenging due to communication overhead and privacy concerns. To resolve these issues, we introduce a novel framework, called MatchMaker, which extends the 3GPP network sharing model to a cloud-based 5G NR-U system. We define new interfaces and messages for facilitating private NR-U operation over managed and shared network infrastructure. According to MatchMaker, the network manager runs a Graph Coloring Evolution (GCE) algorithm to learn potential interference between operators and match them with channel resources.