David Lynch

David Lynch

PhD Thesis Title: Automated Self-Optimization in Heterogeneous Wireless Communications Networks using Techniques from Evolutionary Computation

Supervisor: Professor Michael O'Neill
External Examiner: Professor Krzysztof Krawiec, Institute of Computing Science, Poznan University of Technology










Abstract

Traditional wireless communications networks cannot scale to satisfy demand during an era of exponentially growing traffic. One solution is to deploy low-powered antennas called Small Cells in traffic hotspots. Small Cells boost capacity by offloading customers from pre-existing Macro Cells. The resulting Heterogeneous Network (or ‘HetNet’) allows both cell tiers to share the same scarce and expensive wireless spectrum. However, inter-cell interference and load balancing issues constrains the capacity of a HetNet. Resolving these optimisation challenges is crucial in the highly competitive trillion dollar telecommunications industry.

In this thesis, a unified, flexible, and fully automated approach for optimising multi-layer HetNets is presented. The proposed approach instruments a powerful nature-inspired technique called evolutionary computation. The following research questions emerge in this context:

1.     How can we design better evolutionary algorithms for this instance of a dynamic and uncertain environment?

2.     What is the best approach for jointly optimising multi-layer HetNets?

3.     How can we manage fairness trade-offs through the use of evolutionary algorithms?

Evolved controllers boost downlink rates by over 220% compared to non-adaptive baselines. Furthermore, the flexible evolutionary paradigm enables the creation of tailored service plans, which will be key to effectively monetising HetNets. The experiments illustrate the suitability and potential of evolutionary computation in managing future wireless communications networks.

 

Discover our Rankings and Accreditations