Wireless Communication with Unmanned Aerial Vehicles : Design Tradeoffs and Machine Learning Techniques
Files in This Item:
|Request a copy||104583561.pdf||2.64 MB||Adobe PDF|
|Title:||Wireless Communication with Unmanned Aerial Vehicles : Design Tradeoffs and Machine Learning Techniques||Authors:||Fontanesi, Gianluca||Permanent link:||http://hdl.handle.net/10197/12946||Date:||2022||Online since:||2022-06-30T13:33:28Z||Abstract:||Unmanned Aerial Vehicles (UAVs) are expanding rapidly in a wide range of wireless network applications. UAVs have inherent mobility, agility, ability to form Line of Sight (LoS) radio links in highly dynamic environments that are promising features in all their potential application scenarios. UAVs are poised to become a key platform in wireless systems and are expected to play an important part as cellular-connected User Equipments (UEs). In addition, UAV can overcome the static nature of ground Base Stations (BSs) deployment to cater to certain Fifth Generation (5G) use cases, such as surging traffic demands in hotspots (stadium, event centres), as well as their availability in emergencies (e.g. natural disasters), where the infrastructure could itself be compromised. UAVs require continuous and reliable connectivity, efficient network and well-designed path planning to fulfil their potential to support future wireless communication. This thesis investigates the interaction between conventional mathematical methods and Machine Learning (ML) techniques, particularly Reinforcement Learning (RL), as a promising synergy for solving various UAV challenges, resulting in increased reliability, throughput and robustness. The first part of this thesis explores the performance of the ground to air link to a UAV-Base Station (BS)/Remote Radio Head (RRH). Stochastic geometry allows us to perform an outage analysis of the Fronthaul (FH) link in an urban scenario for different UAV heights and blockages. In addition, we evaluate the ability of the Backhaul (BH)-FH link to satisfy a target rate on the access link. We extend the evaluation to a Search and Rescue scenario over the sea through computer-based simulations. The results of these two studies show the potential bottlenecks for the ground to air BH/FH and offer insights on how to improve the resilience and create a reliable high capacity ground to air link. To this aim, in the second part of this dissertation, we explore the use of Machine Learning (ML) techniques, especially Deep Reinforcement Learning (DRL). The complex task of UAV trajectory design in communication scenarios is generally a non-convex optimization problem. DRL is a framework able to tackle and automate it. We apply DRL to jointly optimize the UAV flying time and the communication performance of the ground to air link. Initially, we solve the communication-aware trajectory design by proposing 3D trajectory and band switch DRL algorithms. Thus, we consider a novel Lyapunov-based Double DQN (DDQN) to solve the trajectory design while ensuring the satisfaction of the connectivity constraint on the ground to air link. While the RL paradigm offers many advantages for solving optimization problems in UAV communication, some practical challenges remain. The ML model training in UAV communication is a costly and time consuming process requiring a high amount of data. Thus, in the last part of this thesis, we address this issue taking advantage of Transfer Learning (TL) approach. In the Transfer Learning (TL) algorithm developed in this dissertation, the agent utilizes a teacher policy. The teacher trajectory policy, previously trained at sub-6 GHz trajectory, is the solution to a connectivity-aware trajectory probilem. Later, the teacher policy’s knowledge gained in a previous domain at sub-6 GHz is transferred into the new domain at millimeter Wave (mmWave) to considerably reduce the demand for expensive UAV flights and data collection. In the last chapter, we discuss possible future research directions and open challenges, such as the computation load distribution and the integration of UAVs with satellite network. The overall contribution of this thesis is the investigation of practical constraints of UAV communication in wireless communication and to propose relative solutions to make UAV communication more reliable and feasible in real scenarios.||Type of material:||Doctoral Thesis||Publisher:||University College Dublin. School of Electrical and Electronic Engineering||Qualification Name:||Ph.D.||Copyright (published version):||2022 the Author||Keywords:||UAV communication; Machine learning; mmWave||Language:||en||Status of Item:||Peer reviewed||This item is made available under a Creative Commons License:||https://creativecommons.org/licenses/by-nc-nd/3.0/ie/|
|Appears in Collections:||Electrical and Electronic Engineering Theses|
Show full item record
If you are a publisher or author and have copyright concerns for any item, please email email@example.com and the item will be withdrawn immediately. The author or person responsible for depositing the article will be contacted within one business day.