Organization:
Funded by: | Incentive Grant EEMCS |
PhD: |  |
Supervisors: | Chair HS:  Daily Supervisor:   |
Collaboration: |  |
Description:
Model-based techniques for system identification, optimization and control are shifting towards data-driven approaches, surged by the outstanding effectiveness of machine learning techniques. When complex, high-dimensional physical systems are involved, data-driven approaches should be blended with rigorous understanding of the physical properties of the considered systems.
An extremely relevant case involves fluid-mechanical systems (in principle infinite-dimensional) which require a model reduction procedure to generate low-dimensional models for fast predictions. Starting from a precise physical description of a fluid mechanical system, our goal is to fit real-world data in a novel physics-aware model-order reduction procedure leading to a model which can be evaluated for simulation and control purposes with unprecedented efficiency.
As a direct consequence, we aim to introduce new tools to accomplish one of the EEMCS missions in one of the most challenging scenarios: energy-efficient robot control in unsteady fluid-structure interaction regimes.