BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Generative Models for Learning Robot Manipulation Skills from Huma
 ns
DTSTART:20171212T100000
DTEND:20171212T120000
DTSTAMP:20260407T020000Z
UID:ae977507c8fe7493e188978d0f5af8731b35c82c06ffcaf6cbcc6a83
CATEGORIES:Conferences - Seminars
DESCRIPTION:Ajay Tanwani\nAbstract:\nA long standing goal in artificial in
 telligence is to make robots seamlessly interact with humans in everyday l
 ife tasks. Robot learning from humans provides a promising route to bridge
  this gap. In contrast to direct trajectory learning from demonstrations\,
  many problems arise in interactive robotic applications that require high
 er contextual level understanding of the environment. This requires learni
 ng invariant mappings in the demonstrations that can generalize across dif
 ferent environmental situations such as size\, position\, orientation of o
 bjects\, viewpoint of the observer\, etc.\n \nIn this talk\, I will expla
 in how we learn invariant representations with a family of hidden semi-Mar
 kov models for segmentation\, recognition and synthesis of robot manipulat
 ion skills. The invariant segments (also termed as sub-goals\, options or 
 actions) adapt the movement of the robot in accordance with the external e
 nvironmental situations such as size\, position and orientation of the obj
 ects in the environment using a task-parameterized formulation. I will add
 ress learning these models in latent spaces to scale them with high-dimens
 ional sensory data and discuss their Bayesian non-parametric extensions un
 der small variance asymptotics for online learning with streaming demonstr
 ations. Practical examples to perform remote manipulation tasks by teleope
 ration over satellite communication will be used to illustrate the perform
 ance of invariant task-parameterized generative mixture models.\n \nBiogr
 aphy:\nAjay Tanwani is a PhD student in the Computer Science department at
  Idiap Research Institute and Ecole Polytechnique Fédérale de Lausanne (
 EPFL)\, Switzerland. He is interested in robotics\, machine learning and a
 rtificial intelligence paradigms for seamless human robot interaction in e
 veryday life tasks. He recently finished his internship with Google X in M
 ountain View\, California. Prior to doctoral studies\, he completed his MS
  in European Masters on Advanced Robotics with Erasmus Mundus Scholarship 
 from University of Genova and Warsaw University of Technology in 2009-2011
 . His noticeable recognitions include Best Manipulation Paper Award Finali
 st at ICRA 2016\, World Summit Youth Award winner 2009\, and winner of sev
 eral all Pakistan robotics and engineering competitions (NERC’07\, SOFTE
 C'08\, COMPPEC'08\, NASCON’08\, IST’05).\n 
LOCATION:BC 410 https://plan.epfl.ch/?room==BC%20410
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR