Internal:HDP-HMM Experiments

From Brown University Robotics

Revision as of 19:34, 19 May 2010 by Jbutterf (Talk | contribs)
Jump to: navigation, search

Contents

Objectives

The objective of this project is produce an ICRA paper that demonstrates the effectiveness of the HDP-HMM model, and multi-valued function regression in general, for robot learning from demonstration. Critical to this goal is producing experiments that demonstrate the advantages of this method for learning from demonstration.

Owner

Jesse Butterfield

Svn Pointer

/pro/rlab/svn/trunk/exp/multiRegressor

Experimental Tasks

Task Data Collection Learned Expert Experiments Video
Ball Tracking ~jbutterf/multiRegressor/HT2F_PaperResults.mat ~jbutterf/multiRegressor/HT2F_PaperResults.mat ~jbutterf/multiRegressor/HT2F_PaperResults.mat
Localize, Reach ~jbutterf/multiRegressor/last1 same same training

results

Grasping Todo Todo Todo
Simple Assembly (One stick, one ball) Todo Todo Todo
Build Gate Todo Todo Todo
Build 3-simplex Todo Todo Todo

Dependencies

This project depends on the MuliRegressor project, rosmatlab, and the nao teleop project Internal:NaoTeleop.