Master Thesis "Meta Learning in Nonlineary Dynamic System Modelling"

The focus of this master's thesis is the investigation of meta learning techniques for fast model adaptation in low-data scenarios. This is especially relevant for reducing end of line commission times for high-mix low-volume systems.

You will support us in the field of data-driven modeling and identification of nonlinear dynamical systems.

You will familiarise yourself with state-of-the-art approaches in transfer learning and meta learning for system identification through a structured literature review.

Under the guidance of our researchers, you will re-implement and evaluate selected reference methods from current research (e.g. in Python or MATLAB) to build a solid methodological foundation.

With the support of our team, you will design and develop a simulation-based validation environment to assess the performance of the implemented methods.

You will analyse and compare the adaptability and efficiency of different approaches on nonlinear system identification tasks.

Optionally, you will apply and validate your methods on a real-world valve test bench to demonstrate practical applicability.

You may publish your results in a scientific journal and present them at a conference.