Learning to Generate Robot Motions from Human Activity Sequences
We study the generation of motions for robot arms and manipulators that are most likely to be appropriate given a particular situation and real-world task constraints, and we will consequently apply these to control humanoid robots.
To this end, we
(1) collect human body motions as trajectories of set of human joints,
(2) use machine learning methods as learning,
(3) develop methods to generate flexible motion of robot arms and
(4) apply them to control in real-world robot control tasks.
This research is supported by the IT R&D Program of MKE/KEIT.
ΆΖ Overall Concept ΆΖ
||Machine Learning for Generating Flexible Motions of Robot Arms
|| May 2011 - April 2013
||Korea Evaulation Institute of Industrial Technology
||Prof. Byoung-Tak Zhang
||Intelligent Autonomous Systems Group (Technische Universitaet Munchen, Prpf.Michael Beetz)
||bjlee at bi snu ac kr