Early Career Spotlight Talk

Pieter Abbeel
University of California at Berkeley
http://www.eecs.berkeley.edu/~pabbeel/

Biography: Pieter Abbeel received a BS/MS in Electrical Engineering from KU Leuven (Belgium) and received his Ph.D. degree in Computer Science from Stanford University in 2008. He joined the faculty at UC Berkeley in Fall 2008, with an appointment in the Department of Electrical Engineering and Computer Sciences. His research focuses on robotics, machine learning and control.

Apprenticeship Learning for High-Performance Robot Control

For many robotic systems human operators still significantly outperform autonomous operation. "Apprenticeship learning" is a new approach to high-performance robot control based on learning for control from ensembles of expert human demonstrations. In this talk I'll present the theory of our apprenticeship learning algorithms as well as experimental results on a variety of challenging real-world control problems. Apprenticeship learning has enabled surgical knot-tying, quadruped locomotion across challenging terrains, and by far the most advanced helicopter aerobatics performed by any autonomous helicopter to date, including maneuvers such as chaos, tic-tocs, and auto-rotation landings which only exceptional expert human pilots can fly.

In the latter part of the talk, I will present initial results on a variant of apprenticeship learning that learns "ab initio" based on robot self-experiments. I'll describe preliminary work on self-learning of robot helicopter flight control starting from a very crude helicopter dynamics model. I'll also describe new work in personal robotics, manipulation of deformable objects, and controlling the flight of cyborg beetles.