About the Project
A fully funded PhD opportunity to participate in the world-leading research undertaken by the EPSRC Doctoral Landscape Award at the University of Sheffield.
Supervised by Dr Anton Selivanov and Dr Morgan Jones.
The University of Sheffield is one of the leading Russell Group universities in the UK. We carry out cutting-edge research with strong links to industry. When you enrol to do a PhD with us, you will be working with world-leading academics and have access to top of the range facilities. As a PhD student you will have the opportunity to gain skills not only to conduct research, but also to take your career to the next level, whether you want to stay in academia, go into industry or the public sector, or set up your own company. You will have access to a range of training and support services to help you excel in your studies and beyond.
Project Description
Control systems are the hidden technology that keeps much of modern engineering working. While modern control theory provides sophisticated methods capable of handling uncertainty, disturbances, and nonlinear behaviour, most industrial controllers rely on relatively simple approaches, particularly Proportional-Integral-Derivative (PID) controllers. This is partly because many advanced methods require a sufficiently accurate model to achieve the desired control objectives. Obtaining such a model can be one of the most difficult, expensive, and time-consuming parts of control design, as real engineering systems contain nonlinearities, friction, delays, actuator limitations, and complex interactions that can be difficult to describe mathematically.
Machine learning (ML) offers a potential way to alleviate this modelling bottleneck by learning complex relationships directly from operational data. However, learned models are inevitably imperfect and may be unreliable in regions where little or no data are available. In safety-critical applications such as aerospace, robotics and autonomous systems, it is not enough for an ML-based control to perform well on average. Engineers need to understand how modelling errors affect the behaviour of the controlled system and whether these errors can be sufficiently quantified to provide formal safety guarantees.
This project will combine the modelling capabilities of ML with the mathematical guarantees of control theory. The central question is: Can machine learning reduce the modelling effort required for advanced control while providing rigorous guarantees that the resulting controller remains safe and reliable despite modelling errors?
The research will involve the following interconnected stages:
- Safe Controller Synthesis: Develop methods for learning nonlinear system dynamics from operational data and quantifying the errors and uncertainties in the resulting models. These uncertainties will be incorporated into Control Lyapunov Functions (CLFs) and Control Barrier Functions (CBFs) to provide mathematical guarantees of stability and safety. Koopman-based models and neural networks will be investigated as candidate approaches.
- Verification and Numerical Validation: Develop computational methods for verifying the safety and stability of learned controllers. Neural-network verification, reachability analysis, and mathematical bounds on modelling errors will be investigated, with performance assessed in terms of data requirements, computational cost, and the operating region for which guarantees can be established.
- Benchmarking Against Classical Control: Determine whether ML provides a genuine advantage over established approaches based on system identification followed by model-based control, including Model Predictive Control and adaptive control. Comparisons will consider performance, data requirements, computational cost, adaptability, and safety. The project will not assume that ML is always superior; it will identify when learning is beneficial and when conventional methods remain preferable.
- Virtual and Hardware Validation: Test the most promising methods in digital-twin environments using Quanser Interactive Labs and subsequently on physical systems such as a Furuta (rotational inverted) pendulum. Experiments will introduce realistic effects including friction, actuator nonlinearities, sensor noise, processing delays, and system disturbances.
The project will establish a principled route from data-driven modelling to safe, advanced control, helping determine when machine learning can genuinely reduce the effort of control design without sacrificing the robustness and reliability expected from control engineering. In the longer term, the resulting methods could enable more capable autonomous systems, robots, aircraft, energy systems and other complex technologies.
How to apply
Interested candidates are strongly encouraged to contact the project supervisors to discuss your interest in and suitability for the project prior to submitting your application.
Please refer to the EPSRC DLA webpage for detailed information about the EPSRC DLA and how to apply. Applicants must ensure that they use the correct application form linked to the EPSRC DLA webpage and must submit an application for October 2027 entry. The form is unique to the EPSRC DLA. It contains the list of advertised projects from which to select, and three questions specific to the DLA.
The application deadline is 17:00 GMT 13 January 2027.
The majority of our DLA studentships will be allocated to candidates who qualify for Home fee status, in accordance with EPSRC funding regulations which place strict limits on international recruitment.
Funding Notes
The award will fund the full tuition fee and a maintenance stipend at the UKRI rate (currently £21,805 per annum) for 3.5 years, as well as a research grant to support costs associated with the project
