Research Areas CE
Hierarchical modelling and simulation
The modelling, simulation and optimisation of increasingly sophisticated engineering problems necessitates the construction of a set of several models, each with their own accuracy (fidelity) and computational complexity, serving at different design stages. The combination and interaction of different models is organised within a hierarchical modelling and simulation context.
Uncertainty quantification and optimisation
Geometry, material, operational environmental parameters come with an uncertainty, typically within a predictable range. The effect thereof on the performance, efficiency and reliability of the device needs to be quantified during design. Moreover, designs are optimised on the basis of their numerical models. The combination of both leads to robust designs.
Multiphysical models and co-simulation
Engineering applications cover phenomena of different physical nature (e.g. fluid dynamics, structural dynamics, electromagnetism), typically heavily influencing each other. Their simulation requires multiphysical models accompanied by weak or strong coupling or co-simulation approaches.
Multiscale and multirate methods
Phenomena in engineering applications happen at different scales in space and time. Efficient simulation techniques rely upon multiscale and multirate methods respectively. To that purpose, solutions are iterated between fine and coarse scales in order to both provide an efficient simulation approach and an accurate description of the physical behaviour at all considered scales.
Holistic computational engineering
Modelling, simulation and optimisation has revolutionised engineering design. Instead of a few physical prototypes, chains and families of simulation models represent the first concept, the intermediate designs and finally the entire array of products. The ensemble of simulation models needs to be stored, verified and checked for consistency. Mappings between models and operations on models need to follow structured patterns, not only for building a relational database, supporting the design process, construction and operational life of the device.
Contemporary simulation tasks necessitate the use of high-performance computing systems and the adaptation of algorithms to the available computing infrastructure. Moreover, appropriate use of such systems requires dedicated research on the algorithmic level.
Machine learning and computational robotics
Development of computational approaches to intelligent behaviour and autonomy encompasses methods from both machine learning and computational robotics. Both machine learning-based and analytical robotics methods are used for modelling, simulation and optimisation of robot systems necessary for robot development and deployment. These methods encompass efficient (i) robot forward- and inverse model identification, (ii) low-level control policy acquisition with automatic assessment of the required uncertainty to ensure stability as well (iii) the generation of long-term behavior (such as optimal trajectory segments, modular behaviour composition and computational approaches to planning).