Engineering Optimisation Group

Many engineering problems involve optimisation, i.e. finding a way to adjust designs to maximize or minimize some objectives (e.g. minimising cost) subject to various limitations or constraints.

Optimisation methods are increasingly used in industry and are also a fruitful area of foundational research. 

Distributed optimisation and control, especially of power systems and electronics

Optimisation techniques can be used to design controllers for power systems to stabilize the power system and converge to an optimal operating point.

Applications have included hybrid AC/DC networks, microgrids, unbalanced distribution networks, distributed secondary frequency and voltage control, data-driven control for grid-forming inverters, data-informed predictive control for energy storage systems, etc. 

X-ray coherent diffractive imaging of small crystals

Crystals are packings of objects in a regular array. The continuing development of a new type of X-ray source called the X-ray free-electron laser (XFEL) provides X-rays with unprecedented brightness, coherence and short duration pulses.

These unique properties of XFELs have allowed very small crystals, only a few repeating units across, to be imaged.

Furthermore, the subsequent new signals, resulting from the coherent illumination of the entire crystal, enable the possibility of solving for the structure of the molecules directly from the diffracted intensity itself.

Optimisation techniques are crucial in every step of the process of converting the noisy diffraction data to a clean image of the molecule.

Knowing how a molecule looks from these images can help humanity develop new pharmaceuticals, novel energy sources, and understand life itself.

Bayesian information fusion and inference for decision making

It is hypothesized that all neural processing and action selection can be formulated as the maximization of Bayesian model evidence, or equivalently, the minimization of variational free energy.

Optimisation techniques are essential for developing computationally feasible free energy minimization methods that play a central role in challenging practical applications including Bayesian belief update/propagation,   heterogeneous information fusion and decision making. 

Research

Courses