Type-2 Fuzzy Single and Multi-Objective Optimisation Systems for Telecommunication Capacity Planning

Capacity planning in the telecommunications industry aims to maximise the effectiveness of implemented bandwidth equipment whilst allowing for equipment to be upgraded without a loss of service. The better implemented hardware can be configured, the better the service provided to the consumers can be. Additionally, the easier it is to rearrange that existing hardware with minimum loss of service to the consumer, the easier it is to remove older equipment and replace it with newer more effect equipment. The newer equipment can provide more bandwidth whilst consuming less power and producing less heat, lowering the overall operating costs and carbon footprint of a large scale network. Resilient routing is the idea of providing multiple independent non-intersecting routes between two locations within a graph. For telecommunications organisations this can be used to reduce the downtime faced by consumers if there is a fault within a network. It can also be used to provide assurances to customers that rely on a network connection such as: financial institutions or government agencies. This thesis looks at capacity planning within telecommunications with the aspiration of creating a set of optimisation systems that can rearrange data exchange hardware to maximise their performance with minimal cost and minimising downtime while allowing adaptations to an exchange’s configuration in order to perform upgrades. The proposed systems were developed with data from British Telecom (BT) and are either deployed or are planned to be in the near future. In many cases the data used is confidential, but when this is the case an equivalent open source data set has been used for transparency. As a result of this thesis the Heated Stack (HS) algorithm was created which has been shown to outperform the popular and successful NSGA-II algorithm by up to 92 % and NSGA-III by up to 69% at general optimisation tasks. HS also outperforms NSGA-II in 100% of the physical capacity planning experiments run and NSGA-II in 68% of the physical capacity planning experiments run. Additionally, as a result of this thesis the N-Non-Intersecting-Routing algorithm was shown to outperform Dijkstra’s algorithm by up to 38% at resilient routing. Finally, a new method of performing configuration planning through backwards induction with Monte Carlo Tree Search was proposed.