BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Safe Learning in Control – Leveraging Online Data for Performanc
 e with System Theoretic Guarantees 
DTSTART:20150327T100000
DTEND:20150327T110000
DTSTAMP:20260508T135630Z
UID:d87c161ae7bcd26272912994164aa1cf2b6feda76f50bf950ca4215c
CATEGORIES:Conferences - Seminars
DESCRIPTION:Dr. Melanie Zeilinger\, Berkeley\, California\, USA\nBio : Mel
 anie Zeilinger is a Postdoctoral Researcher in a joint program with the Em
 pirical Inference Department at the Max Planck Institute for Intelligent S
 ystems in Tuebingen\, Germany and the Department of Electrical Engineering
  and Computer Sciences at the University of California\, Berkeley. From 20
 11-2012 she was a postdoctoral fellow at the École Polytechnique Fédéra
 le de Lausanne (EPFL) in Switzerland. She received the Ph.D. degree in Ele
 ctrical Engineering from ETH Zurich in Switzerland in 2011\, and the diplo
 ma in Engineering Cybernetics from the University of Stuttgart in Germany 
 in 2006. She conducted her diploma thesis research at the University of Ca
 lifornia at Santa Barbara in 2005-2006. She received the ETH medal for her
  dissertation in 2012 and was awarded a Marie Curie Fellowship for Career 
 Development by the European Commission in 2011. Her research interests rev
 olve around distributed control and optimization\, as well as safe learnin
 g-based control\, with applications to energy distribution and management 
 systems and human-in-the-loop control.\nAbstract: Demanding performance re
 quirements combined with increasing complexity\, uncertainty and human int
 eraction in many emerging application problems\, e.g. in robotic\, transpo
 rtation\, or power systems\, are pushing traditional control methods to th
 eir limits. A new opportunity to address these challenges is offered by se
 nsor technologies with the ability to collect large amounts of data online
 .  While machine learning provides powerful techniques to analyze and uti
 lize such large-scale data\, safety concerns when integrating them in a cl
 osed-loop\, automated decision-making process represent a key limitation f
 or leveraging their potential.\nIn this talk\, we will discuss some of our
  recent work towards an automatic controller synthesis that utilizes onlin
 e data to enhance system performance\, while ensuring satisfaction of safe
 ty conditions at all times.  We show how a predictive controller can be s
 ystematically tailored to the particular system at hand by improving predi
 ctions\, quantifying uncertainties and/or tailoring the objective function
  online based on data\, providing a high performance controller with reduc
 ed development times. Then\, a safety wrapper is introduced that exploits 
 reachability analysis to ensure satisfaction of constraints for any online
  control scheme. The key novelty is the learning capability of the wrapper
  itself\, utilizing data to find the largest region of safe operation wher
 e a performance-maximizing controller can be employed. Finally\, experimen
 tal results are shown for a quad-rotor safely learning to fly.
LOCATION:ME B1 10 http://plan.epfl.ch/?room=MEB110
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR