BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Context-aware controller inference - Benjamin Peherstorfer (Couran
 t Institute of Mathematical Sciences)
DTSTART:20230424T090000Z
DTEND:20230424T100000Z
UID:TALK198391@talks.cam.ac.uk
DESCRIPTION:This work introduces a data-driven control approach for stabil
 izing high-dimensional dynamical systems from scarce data. The proposed co
 ntext-aware controller inference approach is based on the observation that
  controllers need to act only on unstable dynamics to stabilize systems un
 der mild assumptions. This means it is sufficient to learn the unstable dy
 namics alone\, which are typically confined to much lower dimensional spac
 es than the high-dimensional state spaces of all system dynamics and thus 
 few state observations are sufficient to identify them. Numerical experime
 nts demonstrate that context-aware controller inference learns stabilizing
  controllers from orders of magnitude fewer state observations than tradit
 ional data-driven control techniques and variants of reinforcement learnin
 g. The experiments further show that the low data requirements of context-
 aware controller inference are especially beneficial in data-scarce engine
 ering problems with complex physics\, for which learning complete system d
 ynamics is often intractable in terms of data and training costs.
LOCATION:Seminar Room 1\, Newton Institute
END:VEVENT
END:VCALENDAR