Embedded Files
The full-day workshop will be organized into a morning and an afternoon session, split by a lunch break. The morning session provides a tutorial-style introduction and overview of GPs in control. It includes lecture-like presentations as well as hands-on examples. The afternoon session focuses on advanced topics, applications of Gaussian processes, and a series of exciting research questions presented by the contributing speakers.
Tentative Schedule
08:30 - 08:40
Warming up: Welcome, motivation, and introduction
Warming up: Welcome, motivation, and introduction
08:40 - 09:20
From Gaussian distributions to Gaussian processes
From Gaussian distributions to Gaussian processes
[Tutorial] Thomas Beckers
09:20 - 10:00
Regression with Gaussian processes
Regression with Gaussian processes
[Tutorial] Maik Pfefferkorn
Coffee Break 10:00 - 10:30
10:30 - 11:30
Learning-based Modeling and Control of Physical Systems
Learning-based Modeling and Control of Physical Systems
[Tutorial] Thomas Beckers
11:30 - 12:30
Gaussian Processes in Model Predictive Control: Basics, Safety, and Stability
Gaussian Processes in Model Predictive Control: Basics, Safety, and Stability
[Tutorial] Maik Pfefferkorn
Lunch Break 12:30 - 01:15
01:15 - 02:00
Kernel methods for robust control: An approximation theory approach
Kernel methods for robust control: An approximation theory approach
[Spotlight] Colin Jones
02:00 - 02:45
Performance guarantees for Gaussian process online learning control
Performance guarantees for Gaussian process online learning control
[Spotlight] Sandra Hirche / Armin Lederer
In this talk we will present recent results on Gaussian process-based control with performance guarantees for highly uncertain systems. Using a model-based learning control architecture we employ Gaussian Processes to learn the uncertain dynamics. The epistemic model uncertainty due to limited training data is explicitly taken into account, such that data-dependent tracking error bounds can be shown for this control architecture. We address computational challenges of Gaussian process models in control by proposing an efficient approximation, such that performance guarantees are kept. The effectiveness of the online learning control approach is demonstrated in experiments.
Coffee Break 02:45 - 03:15
03:15 - 04:00
Exploiting Gaussian Processes in Predictive Control
Exploiting Gaussian Processes in Predictive Control
[Spotlight] Rolf Findeisen
04:00 - 04:45
Gaussian Processes for Learning and Control:
Opportunities for Real-World Impact
Gaussian Processes for Learning and Control:
Opportunities for Real-World Impact
[Spotlight] Karl Berntorp
04:45 - 04:50
Closing remarks/discussion
Closing remarks/discussion
Page updated
Google Sites
Report abuse