BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Event-triggered sampling for state estimation - Maben Rabi
DTSTART:20100219T140000Z
DTEND:20100219T150000Z
UID:TALK21876@talks.cam.ac.uk
CONTACT:Dr Ioannis Lestas
DESCRIPTION:Wireless networks are replacing Field buses in Discrete manufa
 cturing and Process control because they: \n\n * Avoid wiring costs\, \n *
  Are simpler to install\, \n * Are simpler to maintain\, and\, \n * Potent
 ially\, allow deploying more sensors and actuators.\n\nHowever\, wireless 
 channels have limited bandwidth and fluctuating\ndelivery rates.\n\nBandwi
 dth limits are Packet rate limits. Under such limits\, it is important to 
 communicate measurement samples and control updates at times when such com
 munication has the most value. Hence\, Event-triggered (ET) sampling strat
 egies are more efficient than periodic sampling. When the network delivers
  packets reliably\,\nET schemes are superior because the  _`silence' betwe
 en successive samples encodes useful information_.\n\nThe best sampling sc
 hemes are solutions of Optimal stopping problems. For sampling scalar line
 ar systems over an infinite horizon with a limit on the average sampling r
 ate\, Delta sampling is optimal. For finite horizons\, it is not. Moreover
 \nits behaviour on finite horizons has some counter-intuitive and unexplai
 ned phenomena.  We will also see that\, unlike under periodic sampling\, E
 T sampling for control is more complicated than ET sampling for state esti
 mation.\n\nFluctuating delivery rates cause some samples to be lost. For t
 he state estimation problem\, we will see whether the loss of some samples
  will destroy the superiority of ET sampling. \n
LOCATION: Cambridge University Engineering Department\, LR5
END:VEVENT
END:VCALENDAR