posted on 2025-05-09, 15:08authored bySamoda Lakshitha Okanda Gamage
Exponential growth of the wireless communications demands for efficient use of the limited available radio spectrum for network communications. Current fixed spectrum usage patterns are inefficient where some of the frequency bands are heavily congested while some other bands are rarely utilized. Cognitive radio has been proposed as a solution to the ever-increasing demand for the radio spectrum. Traditionally, channel sensing has been widely deployed by secondary users (SU) to determine the presence of primary users (PU). However, inherent limitations associated with channel sensing may make the secondary resource allocation unreliable and often energy-inefficient. Geolocation databases (GDB) have been proposed as an alternative to channel sensing, to overcome those limitations. This thesis proposes a GDB-based cognitive radio network (CRN) architecture to support different networks with heterogeneous radio access technologies. Novel algorithms have been proposed that exploit features of the GDB architecture to improve the QoS of the SU networks. Performance of the proposed algorithms is further improved by utilizing a Markov-based saturation throughput estimation mechanism and an auto regressive moving average-based traffic peak duration prediction mechanism. Simulation results show that the proposed algorithms yield better QoS levels compared to existing GDB-based resource allocation algorithms. QoS provision in CRNs is further discussed based on statistically guaranteed QoS provisioning. Effective capacity (EC) is utilized as the performance metric where the required QoS levels are expressed by means of the QoS exponent. Maximization of the total EC of a multi-user CRN, while meeting transmit power constraints and PU protection policies is considered. Two successive convex approximation - based algorithms are proposed to solve the non-convex problem in an iterative manner. Using numerical analysis, it is shown that the two algorithms converge to the same optimum value after several iterations irrespective of their staring points. Furthermore, the thesis presents a performance comparison between GDB-based and sensing-based CRN architectures. Using mathematical models, it is shown that a GDB-based CRN can achieve higher SU throughput, lesser interference at the PU and higher energy efficiency compared to a sensing-based CRN.
History
Year awarded
2019.0
Thesis category
Doctoral Degree
Degree
Doctor of Philosophy (PhD)
Supervisors
Khan, Jamil (University of Newcastle); Ngo, Duy (University of Newcastle)
Language
en, English
College/Research Centre
Faculty of Engineering and Built Environment
School
School of Electrical Engineering and Computer Science