BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//wp-events-plugin.com//7.2.3.1//EN
TZID:Europe/Paris
X-WR-TIMEZONE:Europe/Paris
BEGIN:VEVENT
UID:87@cran.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20240328T150000
DTEND;TZID=Europe/Paris:20240328T160000
DTSTAMP:20240318T164050Z
URL:https://www.cran.univ-lorraine.fr/events/seminaire-tyler-westenbroek/
SUMMARY:Seminaire Tyler Westenbroek
DESCRIPTION:- Teams link : https://teams.microsoft.com/l/meetup-join/19%3aM
 VHWcoJpE9ydRWENhbnWqgLENWmTloOdNUbGwTj3dYs1%40thread.tacv2/1703782252879?c
 ontext=%7b%22Tid%22%3a%22158716cf-46b9-48ca-8c49-c7bb67e575f3%22%2c%22Oid%
 22%3a%22955ae724-3c70-4f81-8a92-327e1dc5af57%22%7d\n\n- Titre : Feedback D
 esign Principles for Efficient and Reliable Robot Learning\n\n- Résumé :
  Truly autonomous robots must rapidly adapt their behavior to new operatin
 g conditions. Classical techniques from control leverage first principles 
 models to synthesize robust controllers and efficient adaptation strategie
 s. However\, these frameworks often break down in real-world scenarios whe
 re models can fail. Data-driven techniques from machine learning such as d
 eep reinforcement learning promise to overcome this challenge. In theory\,
  these paradigms have the ability to learn high-performing controllers dir
 ectly from real world data\, eschewing the need for a first-principles mod
 el. However\, in current practice they are too data-inefficient and unreli
 able to see widespread practical deployment. In this talk I will discuss h
 ow to fuse these paradigms in a principled way\, by embedding design techn
 iques from feedback control into reinforcement learning setups. This appro
 ach enables reinforcement learning algorithms to leverage known structures
  in approximate dynamics models\, while maintaining the flexibility to lea
 rn from unmodeled dynamics. First\, I will discuss principled reward shapi
 ng approaches that use feedback design techniques to produce stable contro
 llers. Second\, I will discuss how to codesign feedback controllers and po
 licy gradient algorithms to make learning in the real world efficient and 
 reliable. I will show analytically how these solutions lead to inherent ro
 bustness guarantees\, while empirically reducing the amount of real-world 
 data required by an order of magnitude. Finally I will discuss directions 
 for future work.\n\n- Biographie : Tyler Westenbroek is a postdoctoral sch
 olar at the University of Washington. He completed his Ph.D. in Electrical
  Engineering and Computer Sciences at UC Berkeley in February 2023\, under
  the supervision of Shankar Sastry. His work aims to develop scalable data
 -driven tools for controlling complex\, high-dimensional robotic systems i
 n the real world. His work leverages techniques from control theory and ma
 chine learning\, and has appeared in top conferences and journals across r
 obotics\, control and machine learning.
CATEGORIES:Département CID,Séminaire projet MODELE
LOCATION:CRAN - ENSEM\, 2\, Avenue de la Foret de Haye\, Voandoeuvre-les-Na
 ncy\, 54516\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=2\, Avenue de la Foret de H
 aye\, Voandoeuvre-les-Nancy\, 54516\, France;X-APPLE-RADIUS=100;X-TITLE=CR
 AN - ENSEM:geo:0,0
END:VEVENT
BEGIN:VTIMEZONE
TZID:Europe/Paris
X-LIC-LOCATION:Europe/Paris
BEGIN:STANDARD
DTSTART:20231029T020000
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
END:STANDARD
END:VTIMEZONE
END:VCALENDAR