BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//talks.cam.ac.uk//v3//EN
BEGIN:VTIMEZONE
TZID:Europe/London
BEGIN:DAYLIGHT
TZOFFSETFROM:+0000
TZOFFSETTO:+0100
TZNAME:BST
DTSTART:19700329T010000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0100
TZOFFSETTO:+0000
TZNAME:GMT
DTSTART:19701025T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
CATEGORIES:Microsoft Research Machine Learning and Perception
  Seminars
SUMMARY:Exploiting Variable Impedance for Robotics: Mimic 
 or Optimize? - Sethu Vijayakumar\, Department of I
 nformatics\, University of Edinburgh
DTSTART;TZID=Europe/London:20120214T140000
DTEND;TZID=Europe/London:20120214T150000
UID:TALK36355AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/36355
DESCRIPTION:It is the year of the London Olympics and appropri
 ately\, this talk is about making robots run faste
 r\, jump higher and throw further. Variable Impeda
 nce refers to the ability to change stiffness and 
 damping during task execution. With novel prototyp
 e robotic actuators capable of fast impedance modu
 lation\, the obvious question is how we can maxima
 lly exploit this capability in an automatic and da
 ta driven manner? In this talk\, I will look at im
 pedance modulation in three different classes of m
 ovements: point-to-point tasks like reaching\, exp
 losive movement tasks like throwing and rhythmic m
 ovement tasks such as walking and running. I will 
 describe an optimal control based formulation of o
 ptimizing both the temporal profile of movement an
 d impedance modulation in a way that is tuned to t
 he dynamics of the plant. Several hardware tests w
 ill serve to highlight the benefits. Further\, I w
 ill illustrates the pitfalls of naively mimicking 
 impedance profiles across heterogeneous systems (e
 .g.\, human limb to VS joints or MACCEPA actuators
 ) and describe a framework that is capable of abst
 racting out the specific plant dynamics while ensu
 ring task optimality. This talk will draw upon con
 cepts of optimal feedback control\, apprenticeship
  learning and model free reinforcement learning be
 sides fundamentals of dynamics representation and 
 learning.
LOCATION:Small public lecture room\, Microsoft Research Ltd
 \, 7 J J Thomson Avenue (Off Madingley Road)\, Cam
 bridge
CONTACT:Microsoft Research Cambridge Talks Admins
END:VEVENT
END:VCALENDAR