Research Associate in "Adaptive data-driven predictive control: theory and algorithms" (m/w/d)
Contributed by: Naim Bajcinca, mec-apps@mv.uni-kl.de
Research Scope
When it is infeasible to derive a mathematical model for a physical process using first principles, we use measured data to model the process. This is called system identification and has been the subject of research for many decades. Once we have the model for the process, we can use it to design the control laws. This is called model-based control. Although model-based control has been quite successful, it has some shortcomings. As discussed, it is a two-step process: first, identify the model and then design the control laws. Both steps typically require optimization tasks. Thus, the overall solution to the control problem may be suboptimal. To circumvent this issue, recently data-driven control approaches have been proposed. In this paradigm, control laws are designed directly from measured data without explicitly identifying the model beforehand.
Research Task / Work Description
Data-driven control is a rapidly evolving paradigm, where control strategies are designed directly from measured data. One of the main approaches in this paradigm seems to be inspired by a key result known as the fundamental lemma of the behavioral system theory, in which a model is defined as a set of trajectories. According to this result, one can generate all input/output trajectories of a controllable linear time-invariant system if one can measure a persistently exciting input/output trajectory of the system. In this project, which is funded by DFG, we will explore data-driven methods to tackle several control problems:
- To derive data-driven necessary and sufficient criteria for structural properties such as controllability and observability of different classes of systems.
- To develop behavioral system theory for nonlinear/stochastic systems. In particular, to generalize Willems et al. fundamental lemma to certain classes of nonlinear deterministic/stochastic systems.
- To develop data-driven predictive control (DPC) algorithms as motivated by the approach of model predictive control (MPC), which is a powerful technique to design optimal control laws for complex tasks. While MPC requires an explicit model description by dynamic equations for its implementation, the aim is to design the predictive control laws based directly on measured data, thus avoiding the need for explicit identification of the underlying model.
- The latter algorithms should be extended to adaptive DPC. Various disturbance models need to be synthesised to guarantee the necessary persistence of excitation. In this context, the adaptation laws for Hankel matrices, lying at the core of data driven modelling should be studied, too.
Qualification
- Above-average Master’s degree in electrical engineering, mathematics, or related field.
- Advanced control engineering and numerical optimization skills beyond the content of basic lectures.
- Familiarity with the behavioral system theory is advantageous.
- Highly motivated, eager to work within a team or independently
- Knowledge of at least one programming language: Matlab, Julia, or Python is expected.
- Organizational and collaborative skills with research partners from different disciplines.
- Proficiency in English and / or German is essential.
Application procedure and deadline:
Applications must include the following elements (as a single PDF file):
- Cover letter with a brief description of why you want to pursue research studies, about what your academic interests are, and how they relate to your previous studies and future goals
- CV including your relevant professional experience and knowledge
- Copies of diplomas and grades from previous university studies
- Two references
- List of publications
Send an email with the required documents to the address: @mec-apps@mv.uni-kl.de. The application deadline is 15. May 2023.
For more information, visit our webpage https://www.mv.uni-kl.de/mec/open-positions/adaptive-data-driven-predictive-control-theory-and-algorithms-with-application-to-process-engineering .
