Multi-species evolutionary algorithms for complex optimisation problems

Lu, Xiaofen (2018). Multi-species evolutionary algorithms for complex optimisation problems. University of Birmingham. Ph.D.

Preview

Lu2018PhD.pdf
Text - Accepted Version
Available under License All rights reserved.
Download (2MB) | Preview

Preview

Lu2018PhD_Erratum_Notice.pdf
Text - Supplemental Material
Available under License All rights reserved.
Download (105kB) | Preview

Abstract

Evolutionary algorithms (EAs) face challenges when meeting optimisation problems that are large-scale, multi-disciplinary, or dynamic, etc. To address the challenges, this thesis focuses on developing specific and efficient multi-species EAs to deal with concurrent engineering (CE) problems and dynamic constrained optimisation problems (DCOPs). The main contributions of this thesis are:

First, to achieve a better collaboration among different sub-problem optimisation, it proposes two novel collaboration strategies when using cooperative co-evolution to solve two typical kinds of CE problems. Both help to obtain designs of higher quality. An effective method is also given to adjust the communication frequency among different sub-problem optimisation.

Second, it develops a novel dynamic handling strategy for DCOPs, which applies speciation methods to maintain individuals in different feasible regions. Experimental studies show that it generally reacts faster than the state-of-the-art algorithms on a test set of DCOPs.

Third, it proposes another novel dynamic handling strategy based on competitive co-evolution (ComC) to address fast-changing DCOPs. It employs ComC to find a promising solution set beforehand and uses it for initialisation when detecting a change. It is shown by experiments that this strategy can help adapt to environmental changes well especially for DCOPs with very fast changes.

Type of Work:

Thesis (Doctorates > Ph.D.)

Award Type:

Doctorates > Ph.D.

Supervisor(s):