UAV-Enabled Wireless Communications: Deployment, Optimization, and Analysis

Abstract

Unmanned aerial vehicles (UAVs), known as drones, are a promising solution, as aerial base stations (BSs) or relays, in wireless communications systems. Due to their high likelihood of line-of-sight (LoS) links and ease of deployment, they play a crucial role in providing faster and better wireless network access and service where extra network resources are needed short term, like sports events, or to those emerging services that require high-capacity communications. Moreover, they can help extend wireless coverage to locations deprived of end-to-end wireless communications services in remote/rural areas due to natural disasters or being distant from the conventional terrestrial BSs. However, utilizing this new technology comes with its own novel challenges. In this dissertation, we focus on unprecedented challenges in UAV communications and networks, considering some unique features of UAV networks, such as their optimal placement and wireless backhaul links. First, we focus on provisioning wireless coverage to those emerging services, like extended reality, demanding high-capacity communications. High frequencies, i.e., millimeter-wave (mmWave) and, terahertz frequency bands offer the substantial bandwidth required for such services. These high-frequency communications, however, depend critically on maintaining LoS connections to user terminals. In practical scenarios, users distributed in three-dimensional space often experience severely limited visibility due to environmental obstructions like buildings and foliage. We study the problem of finding an optimal 3D placement and antenna orientation for mmWave-equipped UAVs to minimize the number of required UAVs while maximizing the signal-to-noise ratios (SNRs) to all users. Our approach formulates this as an integer linear programming (ILP) optimization problem, establishes its computational intractability (NP-hardness), and develops a computationally efficient geometric algorithm that consistently achieves near-complete LoS coverage across diverse simulation scenarios. Our second research thrust targets wireless connectivity in remote rural environments—such as agricultural Internet of Things (IoT) deployments—where conventional terrestrial infrastructure is limited or absent. A fundamental challenge in such UAV-assisted networks is determining the minimal UAV deployment that simultaneously achieves two objectives: complete ground user coverage and reliable wireless backhaul connectivity linking all UAVs to terrestrial BSs. We formulate this joint optimization—termed the Backhaul-and-coverage-aware Drone Deployment (BoaRD) problem—as an ILP problem and prove its NP-hardness. Our solution approach employs a graph-theoretic algorithm that efficiently solves the problem with provable performance bounds. Comparative analysis using ILP solvers demonstrates that our algorithm achieves near-optimal performance for smaller problem instances. For large-scale scenarios with extensive coverage areas and numerous users, comprehensive simulations show our algorithm substantially outperforms baseline algorithms while guaranteeing complete user coverage and end-to-end connectivity. Finally, building upon these deployment optimization contributions, our third research thrust develops a comprehensive analytical framework for multi-hop UAV-assisted cellular networks. While the previous work provides deterministic algorithms for specific deployments, understanding system-wide performance requires statistical modeling of networks with random spatial distributions. We develop a comprehensive stochastic geometry framework for analyzing multi-hop UAV-assisted cellular networks that addresses fundamental gaps in existing analytical approaches. Traditional stochastic geometry techniques for terrestrial networks are insufficient for characterizing the complex 3D spatial relationships, interference patterns, and unique propagation characteristics inherent in multi-hop UAV deployments. We extend existing mathematical frameworks to accommodate the distinctive features of aerial networks, including realistic 3D spatial distributions of UAVs across multiple operational altitudes, probabilistic air-to-ground channel models that distinguish between LoS and NLoS conditions, and the intricate interference correlations that arise in multi-hop communication paths. Our framework derives novel mathematical constructs and probability distributions that enable precise characterization of multi-hop network behavior under random spatial deployments in the 3D space. We provide comprehensive closed-form expressions for coverage probability analysis covering both amplify-and-forward (AF) and decode-and-forward (DF) relaying protocols, accounting for the hybrid communication scheme where UEs can connect either directly to serving BSs or through the multi-hop UAV network based on received signal quality. Additionally, we introduce optimal relay selection strategies that maximize end-to-end SINR by jointly considering all link qualities in the formed multi-UAV network and accounting for the complex interdependencies between sequential links in the presence of interference. Through extensive theoretical analysis and simulation validation, our results demonstrate that well-designed multi-hop UAV networks can significantly enhance coverage probability and network reliability compared to single-hop architectures, particularly in challenging environments where direct links between UAVs and terrestrial BSs are weak or unavailable due to distance or environmental obstructions.