Seminar: Learning-based Control

  The Furuta Pendulum and  one of our scientists in the back Copyright: © Wolfram Scheible / MPI-IS
 

Machine learning and artificial intelligence have a huge potential to enable autonomous robots, vehicles, and smart infrastructure. However, important research challenges have to be overcome before autonomous systems can operate in the real world. In particular, we need algorithms that can provide safety guarantees, are resource-efficient, and can effectively combine prior knowledge with data.

 

Facts

  • Cycle: winter semester
  • Assessment: written homework and presentation
  • Credits: 4 CP
  • Language: English
 

Learning-based control is a recent and very active area of research that addresses these challenges and broadly denotes the intersection of the areas of automatic control and machine learning.


In this seminar, you will study recent research papers on topics such as Gaussian processes for dynamics and control, (deep) reinforcement learning, hybrid physics- and data-based modeling, and verification for learning-based control. This is an interdisciplinary seminar offered in Mechanical Engineering and Computer Science.