Data Driven Control
Learning-based and data-driven techniques are currently revolutionizing how we model, predict, and control complex dynamical systems. The most pressing scientific and engineering problems of the modern era are not amenable to empirical models or conceivable from first-principles. With that in mind, researchers are turning to learning-based approaches for a diverse range of complex systems, such as turbulence, the brain, climate, epidemiology, finance, robotics, and autonomy. These systems are typically nonlinear, dynamic, multi-scale in space and time, high-dimensional, with dominant underlying patterns that should be characterized and modeled for the eventual goal of sensing, prediction, estimation, and control. With modern mathematical tools, enabled by unprecedented availability of data and computational resources, we are now able to tackle those problems directly from measured data.
In this course, we focus on the state-of-the-art of data-driven approaches. In particular, we will focus on the key challenge of designing controllers directly from measured data. Numerically reliable algorithms will be demonstrated in MATLAB.
- Recap of model-based control
- Linear system identification
- Fundamental Lemma
- Data-driven simulation
- Data-driven stabilization
- Data-driven LMI based controller design
- Data-driven predictive control
- Dynamic mode decomposition
- Koopman operators
- Neural networks and dynamical systems
- Deterministic and stochastic numerical methods for system solving
- Universal approximation theorems.
- Course script "Data driven Systems and Control" written by N. Bajcinca, M. Mukherjee, V. K. Mishra, S. Hiremath, 2022.
- S. L. Brunton and J. N. Kutz: "Data-driven science and engineering: Machine learning, dynamical systems, and control", Cambridge University Press, 2019.
- P. Deuflhard and P. Bornemann: "Scientific computing with ordinary differential equations", Springer-Verlag, Berlin, Heidelberg, 2002.
- C. De Persis and P. Tesi: "Formulas for data-driven control: Stabilization, optimality, and robustness", IEEE Transactions on Automatic Control, 65, 909-924, 2019.
Termine: Donerstag, 11:45 – 13:15 Uhr
Written Exam: 120-150 Min.
Credit Points: 3ECTS
Basic Knowledge of MATLAB
Lecture slides : OLAT