Distributed Embedded Intelligence

Development of fast numerical optimisation methods based on operator splitting for nonlinear model predictive control; parallelisable algorithms and Embedded GPU implementations for large-scale optimisation problems such as stochastic and risk-averse optimal control problems; applications of optimisation in signal processing such as recursive compressive sensing and image restoration.

Intelligent Uncertain-Aware MPC

Stability theory for uncertain systems with Uncertain Uncertainty (inexact knowledge of the underlying probability distribution) by leveraging results from the fascinating theory of risk measures; development of control schemes for safety-critical systems such as Cobotics.

Intelligent cognitive autonomous vehicles and robots

Model predictive control can be a key enabler for autonomous vehicles, robots and cobots. The most challenging questions in this endeavour are (i) MPC formulations, more often than not, lead to nonconvex formulations; how can we derive simple such formulations, which can be solved efficiently online? (ii) in particular, what is the best way to model obstacle avoidance constraints? (iii) how can we solve the associated optimal control problems in real time, accurately and fast? and (iv) how should we account for the inevitable uncertainty related to the motion of people, other vehicles, road conditions and other factors?

Embedded Optimisation for Advanced Manufacturing

Advanced and intelligent manufacturing has been identified by both EPSRC and the European Commission as a key emerging technology. Modern manufracturing systems are fast and complex dynamical systems, which involve networks of interconnected devices, combining continuous and discrete components, evolving across multiple (fast and slow) time scales and under high performance requirements in increasingly uncertain contexts.

Model predictive control (MPC) is a successful control methodology that originated from the process industry, but is increasingly being used for fast and complex industrial dynamical systems. Pertinent problems are often ill-conditioned and must be solved within stringent runtime limits. Parallelisation on GPUs or FPGAs can be exploited to solve such large-scale problems. Quasi-Netwonian and semismooth Netwonian algorithms can play a key role in dealing with ill conditioning.

Smart Infrastructure Networks

Model predictive control is nowadays the method of choice for the control of water and power networks. Yet, there remain open questions related to how to deal with the underlying uncertainty. In fact, with the deregulation of energy markets, this uncertainty should only be expected to become a more important factor.

Biomedical and Health care applications

Optimal drug administration and model predictive control for fractional-order pharmacokinetics; MPC embedded in wearable or integrated devices such as artificial organs; Machine learning methods for predictive toxicology and drug discovery

Postgraduate level

• Control and Estimation Theory (2021-): the module focuses on optimal control, model predictive control and its stability/invariance properties, and optimal estimation with emphasis on the Kalman filter and its variants

Undergraduate level

• Signals and Control (2022-): The control part of this module covers a range of standard topics in classical control theory: the Laplace transform and its inverse and frequency-domain methods for the analysis of linear dynamical systems and design of PID controllers
• Circuits and Control (2019-2022)

Textbook

The textbook "Control systems: an introduction" was recently published. This book’s objective is to equip the students of engineering schools with the necessary theoretical tools and programming skills to analyse dynamical systems and design appropriate controllers.