DOI
https://doi.org/10.25772/7H29-AF98
Defense Date
2023
Document Type
Dissertation
Degree Name
Doctor of Philosophy
Department
Computer Science
First Advisor
Tamer Nadeem
Abstract
The emergence of machine learning approaches for wireless communication protocol design has become a key paradigm for future wireless communication and systems, particularly for the fifth and sixth generations of mobile communications. These approaches are essential to improve resource management, networking, mobility management, etc., to accommodate the growing needs for data traffic. However, there are limitations to improving algorithmic approaches due to device heterogeneity, service diversification, dynamic network environment, network scale, and the growth in the amount of data related to applications, users, and networks. This thesis proposes novel learning-driven frameworks to design communication protocols that can learn how to decide near-optimal protocols in different environmental contexts, such as device characteristics, application requirements, user objectives, and network conditions. It presents the feasibility of these approaches through simulation, emulation, implementation, and experimental evaluation in different realistic settings, not only by simulation but also by experimental evaluation in practical networks for multi-user mobile networks. Additionally, this research describes a cross-layer dual-phase learning-driven approach to show the feasibility of combining lower-layer multi-user MAC capabilities with upper-layer application requirements, such as multi-user video streaming, to exploit an unexplored cross-layer approach. Overall, this thesis builds a foundation for on-device learning-assisted communication protocol design, shifting from rule-based protocol design to the design and development of self-driven protocols for next-generation networks. The proposed frameworks have shown great potential to realize orders of magnitude increase in data rates, decrease in delay, and protocol's robustness, which could provide new insights into protocol design optimization.
Rights
© The Author
Is Part Of
VCU University Archives
Is Part Of
VCU Theses and Dissertations
Date of Submission
5-12-2023