Research Projects

Main content

MPC with Gaussian Processes

 

Predictive control requires accurate system models for high performance and constraint satisfaction. By using machine learning tools we develop techniques that can improve model accuracy during closed-loop control while maintaining safety guarantees. more...

User feedback-based control

 

User feedback offers potential for ensuring user satisfaction by automatic and online adjustment of a controller in interaction with, or acting in the environment of a user. Our learning-based control solutions utilize user feedback in order to match user-specific demands and address the predicament of trading-off exploration and exploitation. more...

Distributed Safe Learning

 

Exploration is necessary when coupled subsystems are learning their behaviours and couplings, but it may lead the coupled subsystems to unsafe scenarios. We seek to exploit the already learned information about the systems to provide safety during exploration.more...

                

Plug and Play Load Control

 

Demand-side management will play a key role for improving reliability and efficiency in an increasingly complex grid balancing large demands, e.g., caused by electric vehicles. We develop techniques that automatically compute optimal load profiles, taking into account network and user requirements, and most importantly varying load connectivity.more...

 
 
Page URL: http://www.idsc.ethz.ch/research-zeilinger/research-projects.html
26.04.2017
© 2017 Eidgenössische Technische Hochschule Zürich