The science of autonomy: A "happy" symbiosis between learning, control and physics.

Series: 
GT-MAP Seminars
Friday, March 9, 2018 - 15:00
1 hour (actually 50 minutes)
Location: 
Skiles 006
,  
GT AE
Organizer: 
In this talk I will present an information theoretic approach to stochastic optimal control and inference  that has advantages  over classical methodologies and theories for decision making under uncertainty.  The main idea  is that there are certain connections between optimality principles in control and information theoretic inequalities in statistical physics that allow  us to solve hard decision making problems in robotics, autonomous systems and beyond. There are essentially two different points of view  of the same "thing" and these two different points of view  overlap   for a fairly general class of dynamical systems that undergo stochastic effects.  I will also present a holistic view of autonomy that collapses planning, perception and control into one computational engine, and ask questions  such as  how organization and structure relates to computation and performance. The last part of my talk includes computational frameworks for uncertainty representation   and suggests ways to incorporate these representations within decision making and control.