Living Labs

Modelling Lab

Our EasyBEM library implements numerical methods for acoustics, electromagnetism and heat transfer, including problems in unbounded domains. We develop boundary element methods (BEM) and a novel operator preconditioner suitable for GPUs.

The lab reports performance beyond other available BEM software with this approach.

We provide advanced training in molecular modelling and simulation, including quantum approaches. Our research covers the electronic structure and dynamics of molecular complexes, with applications including low-temperature plasma physics.

Application of discrete optimization techniques to load balancing, scheduling and planning, including graph-based approaches to parallelization of computations.

Our research addresses geotechnical inverse problems and uncertainty modelling using Bayesian methods and physics-informed neural networks. These are areas for research cooperation with partners facing complex modelling problems.

Mathematical modelling and optimal scheduling, drawing on expertise in graph structures, optimization and parallel computation.

We develop smooth and nonsmooth optimization methods and software for shape optimization, flow and contact problems, and robotics, including a nonsmooth optimization library.

Our optimal control methods and software support real-time collaborative robotics, including robotic arms, Bennett mechanisms, humanoids and drones.

We develop quadratic programming (QP) solvers and FETI domain decomposition methods, including massively parallel implementations for demanding engineering problems such as contact mechanics.

Developed with the Institute of Geonics, the open-source PERMON library extends PETSc with QP solvers, FETI methods and support vector machine (SVM) implementations. It has attracted interest from teams at Argonne and Oak Ridge National Laboratories and has been used for automated forest-fire detection from NASA satellite imagery.

We combine heterogeneous datasets from different sources, often incorporating specific expert information.

Applications include extracting information from DNA microarray chip images, reconstructing three-dimensional blood-vessel models from CT scans, and estimating or predicting the states of dynamic systems, such as the position of a robot from sensor data.