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:CUED Control Group Seminars
SUMMARY:Embedded Optimization for Optimal Control of Mecha
tronic Systems - Professor Moritz Diehl (K.U. Leuv
en)
DTSTART;TZID=Europe/London:20100507T140000
DTEND;TZID=Europe/London:20100507T150000
UID:TALK21505AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/21505
DESCRIPTION:Many branches of engineering employ linear mapping
s between some input and\noutput sequences\, most
prominently in control engineering and in signal\n
processing. Examples are PID or other linear contr
ollers\, the Kalman\nFilter\, as well as the many
filters used in sound processing e.g.\nin loudspe
akers or hearing aids. These linear maps are usual
ly only useful\nfor one special set of conditions\
, when no constraints are violated\, while\nthey n
eed to be adapted whenever the conditions change.\
n\nA completely different approach is the followin
g: we generate a map\nbetween inputs and outputs\n
via embedded optimization\, i.e. the outputs are g
enerated as the solution\nof parametric optimizati
on problems that are solved again and again\, each
time\nfor different\nvalues of the input\nparamet
ers. This approach directly generates a nonlinear
map between\ninputs and outputs\, and allows to ea
sily incorporate constraints and user\ndefined obj
ectives. It can be shown that this\napproach is ab
le to generate any continuous input-output map eve
n if we\nrequire the optimization problems to be c
onvex in both inputs and\noutputs\, which is the m
ost favourable case [1].\n\nThe structure of the e
mbedded optimization problems needs to be exploite
d\nto the maximum\, as many applications require s
ampling times in the order\nof milli or even micro
seconds. We present four structure exploiting\nalg
orithms that were used in applications:\n\n(a) a c
onvex time transformation for time optimal robot a
rm control [4]\n\n(b) online active set strategy f
or an optimal pre-filter for machine tools\n[3]\n\
n(c) nonlinear real-time iterations for model pred
ictive control of power\ngenerating kite systems [
2]\n\n(d) a duality and Fourier based approach to
optimal clipping in hearing\naids.\n\nThe talk wil
l present joint work with J. Swevers\, M. Moonen\,
J. De\nSchutter\, T. Van Waterschoot\,\nL. Vanden
Broeck\, D. Verscheure\, B. Houska\, H.J. Ferreau
\, and B. Defraene.\n\nReferences\n\n[1] M. Baes\,
M. Diehl\, and I. Necoara. Every continuous nonli
near control\nsystem can be\nobtained by parametri
c convex programming. IEEE Transactions on Automat
ic\nControl\,\n53(8):19631967\, September 2008.\n\
n[2] A. Ilzhoefer\, B. Houska\, and M. Diehl. Nonl
inear MPC of kites under\nvarying wind conditions\
nfor a new class of large scale wind power generat
ors. International\nJournal of Robust and\nNonline
ar Control\, 17(17):15901599\, 2007.\n\n[3] L. Van
den Broeck\, M. Diehl\, and J. Swevers. Embedded
optimization for\ninput shaping.\nIEEE Transaction
s on Control System Technology\, 2009. In press.\n
\n[4] D. Verscheure\, B. Demeulenaere\, J. Swevers
\, J. De Schutter\, and M. Diehl. Time-optimal\npa
th tracking for robots: a convex optimization appr
oach. IEEE\nTransactions on Automatic Control\, 20
09. Accepted for publication.\n
LOCATION: Cambridge University Engineering Department\, LR5
CONTACT:Dr Ioannis Lestas
END:VEVENT
END:VCALENDAR