Dynamic multi-objective optimization using evolutionary algorithms

The goals of this project are to:

• Expand on current understanding of properties, features and parameter impacts of dynamic optimization problems with multiple objectives.
• Use realistic motivations and characteristics, define novel test problems for the combinatorial domain within dynamic multi-objective optimization.
• Provide insights into the impacts of increasingly complex characteristics from realistic systems including implementing multiple dynamic dependencies within multi-objective benchmark problems.

The details

The project aims to better the understanding of the nature of changes, including the impacts of frequency and magnitude of change events, in dynamic optimization problems. A focus on reproducibility, transparency and ensuring the easy replication of results highlights inconsistent practice in previous research. Addressing a gap in the current available test problems, constructing and investigating mathematical problems that incorporate features observed in realistic systems is another major output of the work.

Graduate researcher profile: Daniel Herring

What did you do before you started your PhD?

I studied from 2013-2017 for an MSci in Natural Sciences from the University of Exeter, UK, with final project entitled, “Using Evolutionary Algorithms to Estimate and Classify Neural Model Parameters from EEG data from Alzheimer’s Patients”. During this time, I took part in the iGEM 2015 Synthetic Biology competition and obtained an EPSRC Vacation Research Bursary to research hybridized meta-heuristics for water distribution networks.

What are the challenges of your research role?

In defining new problems and extending definitions of problem types to include novel characteristics, there are limited resources to draw on in terms of established practice in the literature. Additionally, a key challenge for dynamic multi-objective optimization is the efficient and meaningful representation of performance measurements.

What is the best part of your research role?

It is exciting and rewarding to be exploring new problems, publishing accessible and clear research whilst improving understanding in the area and contributing to the knowledge in the field. Publishing meaningful research and pushing the boundaries of the field is fulfilling and being able to present this work at international conferences is a valuable experience.

Where do you wish to go after your PhD? Do you want to enter industry or continue doing more research in academia?

My ideal pathway would be to continue in research after the completion of my PhD. I would continue my focus on developing understanding of dynamic multi-objective optimization problems further and extend this into the many-objective domain, automated algorithm-selection methods as well as practical application and deployment into real world scenarios including aerospace and transport problems, vehicle fleet logistics and energy network optimization.

Supervision team

• The University of Melbourne: Professor Michael Kirley
• University of Birmingham: Professor Xin Yao