My research is at the intersection of robotics and machine learning.
Specifically, in the area of Human-Robot Interaction I focus on
Human-Robot Policy Transfer, where a human user instantiates, on a
robot, an autonomous control policy to perform a desired task. I
approach this problem from a Learning from Demonstration standpoint,
whereby the control policy is estimated from data representative of
demonstrations of the task. I use teleoperative demonstration to
avoid correspondence issues and frame tasks as learning finite state
machines. Within this context, I have focused on learning the number
of machine states and their individual policies using nonparametric
Bayesian methods.
Recent Activities:
Teaching old dogs new tricks: Incremental multimap regression for
interactive robot learning from demonstration.
Daniel H Grollman.
PhD thesis, Brown University, May 2010.
[ bib |
Text |
Slides ]
Robots, and Learning, and the Future! Oh my!.
Applied Minds, Glendale, CA, July 2009.
[ bib |
Slides ]
Rgame: Embodied gaming for robot learning by demonstration.
IJCAI 2009 Robot Challenge - LbD, Pasadena, CA, July 2009.
[ bib |
Text |
Teaser ]
Contact info:
Daniel Grollman
EPFL, Ecole Polytechnique Federale de Lausanne
EPFL-STI-I2S-LASA, Station 9
CH 1015, Lausanne, Switzerland