zurück zur Liste

Wann?

27. November 2024, 17:15-19:00

Wo?

Hörsaal der Kernphysik
S2|14 24
Schlossgartenstr. 9
64289 Darmstadt

Hörsaal der Kernphysik , S2|14 24 , Schlossgartenstr. 9 , 64289 Darmstadt

Veranstalter

FB Mathematik

giesselmann@mathematik.tu-darmstadt.de

Prof. Dr.-Ing. Rolf Findeisen, TU Darmstadt

Model-based planning and control approaches, such as model predictive control (MPC), are widely applied across various fields, from process industries and robotics to autonomous driving. Their popularity stems from the ability to directly incorporate safety constraints into the optimization process. However, ensuring constraint satisfaction becomes challenging when faced with disturbances, uncertainties, or model variations, as well as issues arising from the numerical solutions of optimization problems.
While data-driven approaches, such as learning system models or disturbances from data, can capture some of these uncertainties, providing stability guarantees remains a difficult task. This talk presents approaches for addressing uncertainties in model predictive control and demonstrates how safety and stability can be maintained rigorously when using Gaussian processes or neural networks to adjust the model and make decisions based on data.

https://www.veranstaltungskalender.tu-darmstadt.de/media/Numerik_1725962556551_255.jpeg
 

Tags

Mathematisches Kolloquium, Mathematik, Numerik