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:Signal Processing and Communications Lab Seminars
SUMMARY:Efficient stochastic optimal control for navigatio
n and motor planning - Bert Kappen\, Radboud Unive
rsity Nijmegen\, The Netherlands
DTSTART;TZID=Europe/London:20070711T110000
DTEND;TZID=Europe/London:20070711T120000
UID:TALK7631AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/7631
DESCRIPTION:This talk discusses a class of non-linear stochast
ic control problems that can be efficiently solved
using a path integral. In this control formalism\
, the central concept of cost-to-go or value funct
ion becomes a free energy and methods and concepts
from statistical physics can be readily applied\,
such as Monte Carlo sampling or the Laplace appro
ximation. \nQualitatively different optimal contro
l strategies for different noise levels can be und
erstood as a result of spontaneous symmetry breaki
ng. When applied to a receding horizon problem in
a stationary environment\, the solution resembles
the one obtained by traditional reinforcement lear
ning with discounted reward. An advantage of the p
ath integral control method over RL is that the co
ntrol can be computed for the current state\, with
out considering all other states and 2) that it ca
n be easily generalized to time-dependent tasks. I
t is therefore a suitable approach for time depend
ent control. We further discuss exploration and an
how agents can approximately compute their coordi
nation using belief propagation.\n
LOCATION:LR6\, Engineering\, Department of
CONTACT:Taylan Cemgil
END:VEVENT
END:VCALENDAR