Josh Bongard, University of Vermont


Josh Bongard, University of Vermont

Sage 4101

September 1, 2010 12:00 PM - 1:30 PM

Industrial robotics has completely transformed heavy industry because such machines can perform the same actions repeatedly in a structured environment, such as a factory floor.  However, we have so far failed to build machines which change their operation on the fly in response to unanticipated environmental or internal change, such as damage.  I will present a class of autonomous robots which we have termed 'resilient machines': machines that constantly build and re-build approximate simulations of themselves and their immediate surroundings, and then use those simulations to internally rehearse behaviors before attempting them in reality.  I will demonstrate how this technique allows a robot to autonomously recover from unanticipated damage; how it can be generalized to teams of robots; and how it can be used in social robotics, in which a robot must find and learn from an appropriate teacher.

Bongard's paper


Add to calendar