Physical dynamic systems learning from interactions with the real world
This talk presents recent algorithms and architectures at the intersection of classical control and reinforcement that allow physical dynamic systems to safely learn from interactions with the real world. The application of these methodologies on real-world mechatronic systems to improve efficiency and performance will be shown.
Practical information:
Want to register?
- Price: free
- Webinars take place once a month. Registration is not necessary, just join the seminar freely.
However, you can leave us your contact information and we will send you a reminder on the day of each seminar.
Georganiseerd door:
Physical dynamic systems like mechatronic systems face increasing challenges with respect to efficiency and quality. Higher levels of adaptivity and automation are required in an industry 4.0 setting. Control strategies based on expert knowledge and physics-based models are often suboptimal as not all physical phenomena might be properly incorporated. Also, the interactions of physical dynamic systems with varying environments bring about uncertainties. Reinforcement learning agents on the other hand allow to learn from interactions but demand expensive and sometime unsafe experimentation.
Teacher / speaker
Guillaume Crevecoeur
Guillaume Crevecoeur is associate professor at Ghent University.
With his team, he conducts research at the intersection of system
identification, control and machine learning for mechatronic and
industrial robotic systems. His goal is to endow physical dynamic
systems with improved functionalities and capabilities when interacting
with uncertain environments, other systems and humans. He is member of
Flanders Make in which he leads the Ghent University activities on
sensing, monitoring, control and decision-making.
Guillaume received his Master (June 2004) and PhD (May 2009) in
Engineering Physics from Ghent University. The focus of his PhD research
was the development of model-based optimization and inverse problem
techniques, mainly for neuroscience applications. After obtaining his
PhD he became a postdoctoral fellow of the Research Foundation Flanders
(FWO) where he put his model-based techniques in a more dynamic context
for real-world applications. In the winter 2011 he was a visiting
researcher at the Technical University Ilmenau and the Physikalische
Technische Bundesanstalt, Berlin, Germany, deepening his knowledge on
optimization and inverse problem techniques.
Since his appointment as associate professor (Oct 2014) within the Department of Electromechanical, Systems and Metal Engineering he has been working on the modelling, optimization and control of mechatronic and industrial robotic systems. He teaches numerical optimization, modelling of dynamical systems, mechatronics and robotics. His research team follows a multidisciplinary approach when closing the loop from sensors to actuators in mechatronic systems to improve and unlock their functionalities and capabilities. This, by advancing upon the hardware design, dynamical system models and control algorithms in which the information world works closely together with the physical world. His research is focused on the synergetic treatment of data and physical knowledge, nonlinear control and machine learning. This to improve the efficiency, performance, robustness, autonomy of safety-critical physical dynamic systems.
Machine Learning in Control and Automation Webinar Series
In this webinar series, we bring together researchers that are interested in, or conducting research on Machine Learning for control and automation purposes. A variety of techniques and their applications will be covered, ranging from traditional machine learning techniques such as System Identification, State Estimation and Model Predictive Control (MPC) to Deep Neural Networks (DNN)-based approaches and Reinforcement Learning. We offer a varied program of national and international speakers. Join us for this one hour webinar series!
Receive updates about this series
Related courses
HPC Café
Seminarie - Online, Karlsruhe - KIT, HammerHAI, Training geselecteerd in de Belgische AI Factory Antenna (AIFA)
