Evolutionary Computation for Dynamic Optimisation in Network Environments








The research on optimisation problems in network environments has a long history but it generally fails to capture real-world scenarios as it usually assumes that both the network environments (such as network topologies, node processing capabilities, interference, etc) and the optimisation problems (such as the user requirements) are known in advance and remain unchanged in the problem-solving procedure. However, most real-world network optimisation problems (NOPs) are highly dynamic, where the network topologies, availability of resources, interference factors, user requirements, etc., are unpredictable, change with time, and/or are unknown a priori. This poses many difficulties for decision makers, generating significant optimisation challenges. This research aims to investigate Dynamic NOPs (DNOPs) in various network environments. The dynamics in both network environments and problems will be studied in depth. DNOPs occur across a wide range of application areas, such as communication networks, transport network, social networks, and financial networks. Our theoretical study in this project will seek fundamental insight that is applicable to multiple application areas, while our applied research will focus on railway networks and telecommunications networks.

Evolutionary Computation (EC) encompasses many research areas, which applies ideas from nature (especially from biology) to solve optimisation and search problems. EC has been successfully applied to many real world scenarios, especially for difficult and challenging problems and those problems that are difficult to define precisely. This project aims to investigate EC methods for solving DNOPs. We aim to gain insight and further our understanding of how different EC methods can be applied to DNOPs via empirical and theoretical studies. It is important to carry out this research at both theoretical and empirical levels, as one can feed into the other. We will work with industrial partners (e.g., Rail Safety and Standards Board, and Network Rail) who will validate our research and participate in our project. We can utilise their skills and expertise in producing the underlying theoretical models, which can then be validated on real-world data supplied by them. This project has great potentials to fundamentally change the way in which DNOPs are treated, both from a real-world point of view and from the point of view of advancing our theoretical understanding. We plan to develop a prototype system, in collaboration with our industrial partners, for our industrial partners.

In order to test and evaluate our newly developed algorithms for DNOPs, we will develop a set of common DNOP models that capture the real-world complexities, and develop advanced EC methods to solve these DNOP models. This will benefit wider research communities due to the ubiquity of DNOPs in so many different fields from communication networks to transport networks to social networks to financial networks. The research results of this project will also be of significant benefit to many industries that involve DNOPs and will provide significant savings both from a cost point of view as well as from an environmental perspective.