Date of Award
Doctor of Philosophy (PhD)
Abstract: The advent of cognitive radio technology has enabled dramatically more options in the use of RF spectrum, allowing multiple transmitters to effectively share spectrum in ways that were previously unavailable (either due to technical limitations or regulatory restrictions). In this dissertation, we investigate approaches to managing RF spectrum use, with a focus on combining multiple control decisions in a mutually beneficial manner.
Our approach to making spectrum management decisions is grounded in Markov decision theory, which has a rich formal foundation and is frequently used to guide decision making in other disciplines. Here, we develop a set of Markov Decision Processes (MDPs) that model the RF spectrum management problem (in various forms). These MDPs are then queried to provide guidance for management decisions, including the combination of both admission and modulation decisions. This results in control decisions that are optimal in expectation.
To address the computational complexity inherent in computing these control decisions, we develop heuristic approaches that mimic the MDP's decisions based upon patterns observed in the MDP decision space. These heuristics are shown to closely approximate the optimal results from the MDP.
Finally, we empirically assess the appropriateness of using Markov decision theory for RF spectrum management by comparing our MDPs to a discrete-event simulation model that relaxes several of the modeling assumptions made in the development of the MDPs.
Paul Min, David Peters, Angela Lee