Abstract
Self-Organising Network (SON) is a dynamic entity with capability of learning and functionality of self-optimisation, which has been desired for various optimisation tasks. If self-optimisation function includes several objectives, Multi-Objective- Optimisation (MOO) methods needs to be carried out; hence a self-optimisation algorithm for such task is ambitious. In this thesis, we introduce an algorithm of self-optimisation multi-objective task using concept of Similarity Measure (SM). The introduced algorithm is applied to concurrent capacity and coverage optimisation in SON use-cases with standard data and is compared with self-optimisation methods in literature for optimisation tasks. Furthermore, a unified framework for performance evaluation in SON is introduced using a Markovian approach. An ergodic Markov model is used to estimate residue Uncertainty ENtropy (UEN) for performance evaluation of underlying SON with a procedural self-optimisation function. A comparison of theoretical results on performance evaluation using Markovian approach is also presented in this study. In addition, a comparison of results is presented in a system level simulation of wireless cellular network for a scenario in SON with hot-spot. The network parameters of antenna and power are used in optimisation with two objectives of users’ throughput and cell fairness as target Key Performance Indicators (KPIs). Finally, in this thesis, we show that the enhancement in measured KPIs, using the introduced self-optimisation algorithm, is Pareto- Koopmans (P-K) efficient in which an efficient transmit strategy is achieved. This efficiency can provide more options to the decision-maker with less conflict problem. Methodology, theoretical approach, analytical evaluation, simulation results and future studies are presented in this thesis.