Abstract
In this century, an explosive growth in the number of end-devices accessing wireless networks triggers several challenges, e.g., high-bandwidth demand, latency-sensitive and computing-hungry tasks, and cloud computing may not be able to fulfill these requirements. The notion of multi-access edge computing (MEC) plays a central role in addressing the above issues. In this context, the present dissertation deals with networked systems including MEC and device-to-device (D2D) communications that can operate in unknown and dynamic environments. The thesis focuses on the theoretical and algorithmic issues at the intersection of optimization, machine learning and networked systems. The objectives and innovative claims are summarized as follows:(T1) Distributed online learning approaches for efficient edge computing and D2D communications; and,
(T2) Collaborative edge caching policies by leveraging machine learning advances.
Optimally allocating computing and communication resources is a critical issue in networked systems. To date, most resource management schemes based on conventional pure optimization approaches usually achieve suboptimal performance. Especially when the environment is dynamic or even unknown, the conventional optimization approaches may no longer be applicable. Typically, solving resource allocation problems requires knowledge of the models that map resource variables to the cost or utility, which may be difficult to model in dynamic and complex environments. Targeting scenarios where models are not available, a model-free online learning scheme is developed in this thesis. Through repeated interaction with the environment, reinforcement learning plays a critical role in optimizing the resource allocation policy.
The overarching objective of this dissertation is to apply state-of-the-art optimization and machine learning tools to the emerging networked systems including edge networks, with the ultimate goal of benefiting daily life.
| Date of Award | 1 May 2022 |
|---|---|
| Original language | English |
| Awarding Institution |
|
| Supervisor | Mohammad Nakhai (Supervisor) & Yansha Deng (Supervisor) |