Data- and Learning-Based Control

Semester: Winter
SWS: Lecture 2 + Exercises 1 + Practical exercises 1, 5 CP

In this course, different data- and learning-based control design techniques are considered. Data-based approaches compute controllers directly from the available input and output data, without the intermediate step of identifying a model of the system. In particular, we will discuss virtual reference feedback tuning, control design based on Willems' fundamental lemma, and the data informativity framework. In learning-based control, some machine learning technique is employed to learn a model of the system (or unknown parts thereof) or directly a suitable controller. Within this course, we will in particular consider approaches from reinforcement learning, using Gaussian Processes, and neural networks.