Distributed Optimisation of LTE Systems

Abstract

This thesis concerns the mathematical modelling of self-organising wireless communication networks.

In particular the thesis considers small cell long term evolution (LTE) wireless communication networks, with femtocells as the devices in mind. Self organisation consists of optimising various aspects of the network in a dynamic way. In this thesis we present a number of novel mathematical analyses, beginning by considering the network interactions and dynamics when three of the self-organisation objectives compete: coverage, quality and energy efficiency. The following chapters then select two of these objectives and develop policies that aid self-optimisation, and we benchmark the performance of these policies. A significant contribution is to show that reinforcement learning can be used to develop policies that require no direct knowledge of a node's neighbours and which outperform our benchmark. It is found that the ability of a node in this network to determine the distance to its nearest neighbours in a distributed way is important for the success of the policies that do not rely on reinforcement learning. Therefore the last part of the thesis develops heuristics that enable the estimation of these distances. In the last chapter the thesis proposes a new and particularly significant improvement (the`Two-Part Pulse Method) to an existing patented method which BT hope to implement.

Date of Award	2019
Original language	English
Awarding Institution	University of Bath
Sponsors	BT Research
Supervisor	Jonathan Dawes (Supervisor), Antal Jarai (Supervisor) & Keith Briggs (Supervisor)