BEGIN:VCALENDAR
CALSCALE:GREGORIAN
VERSION:2.0
METHOD:PUBLISH
PRODID:-//Drupal iCal API//EN
X-WR-TIMEZONE:America/New_York
BEGIN:VTIMEZONE
TZID:America/New_York
BEGIN:DAYLIGHT
TZOFFSETFROM:-0500
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
DTSTART:20070311T020000
TZNAME:EDT
TZOFFSETTO:-0400
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0400
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
DTSTART:20071104T020000
TZNAME:EST
TZOFFSETTO:-0500
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
SEQUENCE:1
X-APPLE-TRAVEL-ADVISORY-BEHAVIOR:AUTOMATIC
UID:251921
DTSTAMP:20261113T140703Z
DTSTART;TZID=America/New_York:20261209T120000
DTEND;TZID=America/New_York:20261209T130000
URL;TYPE=URI:https://www.wpi.edu/news/calendar/events/robotics-engineering-
 colloquium-speaking-series-nathaniel-dennler
SUMMARY:Robotics Engineering Colloquium Speaking Series: Nathaniel Dennler
DESCRIPTION:Optimizing at Deployment-Time from User Interaction\n\n  \n\n\n
 Abstract: For a robot to meaningfully assist people\, the robot needs to u
 nderstand what that person is trying to accomplish and execute the behavio
 rs that help the user achieve that goal. However\, it is challenging for e
 veryday users to express their goals and preferences in a way that is cond
 ucive to the optimization of robot behavior. For example\, users cannot ea
 sily articulate an explicit numerical value that describes how good a robo
 t’s behavior is\, let alone its gradient with respect to the robot’s a
 ctions.\nFortunately\, users naturally produce observable behaviors\, like
  gaze\, speech\, or movement that hint at the user’s underlying goals. M
 oreover\, these behaviors can be intentionally elicited during interaction
  with the robot. My work develops algorithms that allow robots to probabil
 istically estimate user goals from such observable behaviors to support ro
 bot optimization.\nThis talk will cover algorithms that address each stage
  of the optimization problem. First\, I will discuss how robots can aggreg
 ate common behaviors across many users to understand which goals are worth
  pursuing\, defining the input space of the optimization. Next\, I will ex
 plain how robots can infer personalized constraints on top of this space b
 y interacting with individual users. Third\, I will describe how humans ca
 n actively guide robot behaviors during the optimization process at deploy
 ment time. To conclude\, I will discuss future directions for optimizing r
 obot behavior so that robots can more effectively adapt to the people they
  assist.\nBio: Nathan is a postdoctoral researcher at the MIT Computer Sci
 ence & Artificial Intelligence Laboratory working with Prof. Andreea Bobu.
  Nathan’s research develops algorithms and methods that enable robots to
  efficiently adapt to users’ preferences.\nPreviously\, Nathan received 
 a Ph.D. in Computer Science at the University of Southern California\, whe
 re he was an NSF GRFP Fellow advised by Prof. Maja Matarić and Prof. Stef
 anos Nikolaidis. Nathan also holds a B.Eng. in Robotics Engineering and a 
 B.S. in Computer Science from Worcester Polytechnic Institute.
END:VEVENT
END:VCALENDAR
