Use this URL to cite or link to this record in EThOS:
Title: Intelligent radio resource management for mobile broadband networks
Author: Zhao, Qiyang
Awarding Body: University of York
Current Institution: University of York
Date of Award: 2013
Availability of Full Text:
Access from EThOS:
Access from Institution:
This thesis studies intelligent spectrum and topology management through transfer learning in mobile broadband networks, to improve the capacity density and Quality of Service (QoS) as well as to reduce the cooperation overhead and energy consumption. The dense deployment of small cell base stations (BSs) is an effective approach to provide high capacity density access. In the meantime, multi-hop wireless backhaul networks enable highly flexible deployment and self-organization of small cell BSs. A heterogeneous small cell access and multi-hop backhaul network is studied in this thesis as mobile broadband system architecture. Transfer learning is applied to Radio Resource Management (RRM) as an intelligent algorithm to improve the performance of conventional reinforcement learning. In transfer learning, a BS trains its knowledge base relying on knowledge transferred from other related BSs, who are selected using an interference coordination strategy. In a network with static topology, cooperation management is developed to identify the maturity of the knowledge base and control the coordination overhead. It is demonstrated in a multi-hop backhaul network that transfer learning delivers a QoS level that is as high as achieved by a fully coordinated algorithm, but with a very low level of information exchange which is close to a fully distributed algorithm. Transfer learning is also studied in rapidly changeable network architectures to provide reliable communication. It is carried out during the changes of network topology, through mapping the learner’s knowledge base to a prioritized action space with Pareto efficiency. This process assists the BSs to quickly identify and adapt to environment changes, and makes effective decisions. Results show that transfer learning significantly reduces QoS fluctuation during traffic variation and topology changes in a highly dynamic network. Furthermore, a dynamic topology management algorithm is developed to intelligently control the working modes of BSs, based on traffic load and capacity in multiple cells. Topology management is demonstrated to reduce the number of activated BSs with adequate QoS performance provided. Dynamic capacity provision between multiple cells is achieved from transfer learning, which significantly improves QoS and reduces energy consumption.
Supervisor: Grace, David Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available