Project description
With the ever-growing complexity of systems and the concurrent surge of data available for control design, learning controllers directly from data is becoming increasingly appealing for researchers and practitioners. Indeed, data-driven control allows one to skip any explicit identification step, which generally takes the lion’s share of control design time, and it enables to use data directly to fulfill the intended control design objective. Nonetheless, especially when considering complex systems, data-driven design strategies are far from being consolidated, due to the lack of guarantees and an actual understanding of the closed-loop behavior before the controller deployment.
Looking for methodological solutions for data-driven control with an eye to practical applications for the control of high-tech systems, the core challenge of this Ph.D. position is to develop techniques and tools that provide computationally tractable, user-friendly, and performing data-driven control schemes. The intended research focuses on understanding how to leverage priors on the controlled system and the closed-loop system in the direct design process, while minimizing the tuning effort on the user side.
The purpose of the research is to develop prior-based data-driven control schemes that effectively and efficiently allow for the control of nonlinear systems. The Ph.D. project envisions an investigation of the role of structural priors on open-loop and closed-loop dynamics on the effectiveness of data-driven controllers, with the final goal of successfully applying the developed data-driven techniques to the control of benchmark nonlinear systems. The subtasks of the project are envisioned to be carried out also in collaboration with several internationally renowned researchers.
Tasks
- Study of the literature of data-driven control, machine learning for control, and reinforcement learning.
- Development of data-driven techniques for the control of nonlinear systems based on priors on the underlying system, the controller’s structure and/or the closed-loop behavior.
- Development of reinforcement learning techniques that leverage the same priors.
- Analysis of consistency, computational efficiency, and convergence of these techniques.
- Empirical validation of controller quality in terms of closed-loop performance and closed-loop stability in simulation and on suitable experimental case studies.
- Dissemination of the results of your research in international and peer-reviewed journals and conferences.
- Writing a successful dissertation based on the developed research and defending it.
- Assume educational tasks like the supervision of Master students and internships.
Job requirements
We are looking for a candidate who meets the following requirements:
- You are a talented and enthusiastic young researcher.
- You have experience with or a background in systems and control, mathematics, optimization, machine learning, data-driven modelling.
- Preferably you finished a master’s in Systems and Control, Mechanical or Electrical Engineering or (Applied) Mathematics.
- You work well in a team, with an interest towards foundational and methodological research.
- You have good programming skills and experience.
- You have good communicative skills and a cooperative attitude in the work of a research team.
- You are creative and ambitious, hard-working, and persistent.
- You have good command of the English language (knowledge of Dutch is not required).
For more information and to apply, visit our website: https://jobs.tue.nl/en/vacancy/phd-position-in-datadriven-control-of-nonlinear-systems-1011773.html
