BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:When Control Changes the Data: Safety under Interaction-Driven Dis
 tribution Shifts
DTSTART:20261109T140000
DTEND:20261109T150000
DTSTAMP:20261002T031434Z
UID:5ace62371b4ef5e1ccbe514eb5e64cee26c0f62b6f8feac20fc8ae2e
CATEGORIES:Conferences - Seminars
DESCRIPTION:Professor Lars Lindemann \, Assistant Professor for Algorithmi
 c Systems Theory in the Automatic Control Laboratory @ ETH Zürich\nThis s
 eminar is co-sponsored by the IEEE-CSS.\n\nAbstract: \nAccelerated by rap
 id advances in machine learning and AI\, there has been tremendous success
  in the design of learning-enabled autonomous systems in areas such as aut
 onomous driving and robotics. These exciting developments are accompanied 
 by new fundamental challenges that arise regarding the safety and reliabil
 ity of these increasingly complex systems due to imperfect learning\, syst
 em unknowns\, and uncertain environments. Statistical tools for uncertaint
 y quantification have gained popularity due to their ability to deal with 
 these challenges. However\, their guarantees rely on i.i.d. data\, an assu
 mption that is violated when control actions change the underlying data di
 stribution.\n\nIn this talk\, I will provide new insight to design safe co
 ntrollers under distribution shifts using robust conformal prediction (CP)
 . I will begin by advocating for the use of CP due to its simplicity\, gen
 erality\, and efficiency as opposed to existing optimization-based verific
 ation techniques. I will then provide an introduction to CP and summarize 
 existing work that uses CP to design probabilistically safe controllers in
  dynamic environments. Subsequently\, we will look into interactive settin
 gs where the system’s behavior may change the environment's behavior\, a
 nd vice versa. This circular dependency creates an interaction-driven dist
 ribution shift that invalidates existing CP guarantees. To deal with this 
 problem\, we propose an iterative framework that episodically updates the 
 controller while robustly maintaining safety guarantees by quantifying the
  potential impact of a controller update on the environment's behavior. We
  realize this via adversarially robust CP where we perform a regular CP st
 ep in each episode using observed data under the current controller\, but 
 then transfer safety guarantees across controller updates by analytically 
 adjusting the CP result to account for distribution shifts. Lastly\, I wil
 l show how these ideas extend to handling policy-induced distribution shif
 ts that arise when using barrier/Lyapunov functions to control uncertain s
 ystems.\n\nBiography:\nLars Lindemann is currently an Assistant Professor 
 for Algorithmic Systems Theory in the Automatic Control Laboratory at ETH 
 Zürich. From 2023 to 2025 he was an Assistant Professor in the Thomas Lor
 d Department of Computer Science at the University of Southern California.
  Before that\, he was a Postdoctoral Fellow in the Department of Electrica
 l and Systems Engineering at the University of Pennsylvania from 2020 to 2
 022. He received his Ph.D. degree in Electrical Engineering from KTH Royal
  Institute of Technology in 2020. Professor Lindemann's research interests
  include systems and control theory\, formal methods\, machine learning\, 
 and autonomous systems. He is a recipient of a European Research Council S
 tarting Grant and two U.S. National Science Foundation Grants. He also rec
 eived the Outstanding Student Paper Award at the 58th IEEE Conference on D
 ecision and Control and the Student Best Paper Award (as an advisor) at th
 e 60th IEEE Conference on Decision and Control\, and has been a finalist 
 for several other best paper awards.
LOCATION:ME C2 405 https://plan.epfl.ch/?room==ME%20C2%20405 https://epfl.
 zoom.us/j/66361455244
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR