DMCA
Towards spatial perception: Learning to locate objects from vision (2012)
| Venue: | in Postgraduate Conference on Robotics and Development of Cognition (RobotDoC |
| Citations: | 3 - 3 self |
BibTeX
@INPROCEEDINGS{Leitner12towardsspatial,
author = {Jürgen Leitner and Simon Harding and Mikhail Frank and Er Förster and Jürgen Schmidhuber},
title = {Towards spatial perception: Learning to locate objects from vision},
booktitle = {in Postgraduate Conference on Robotics and Development of Cognition (RobotDoC},
year = {2012}
}
Share
 |
 |
 |
 |
||
OpenURL
Abstract
Our humanoid robot learns to provide position estimates of objects placed on a table, even while the robot is moving its torso, head and eyes (cm range accuracy). These estimates are provided by trained artificial neural networks (ANN) and a cartesian genetic programming (GP) method, based solely on the inputs from the two cameras and the joint encoder positions. No prior camera calibration and kinematic model is used. We find that ANN and GP are both able to localise localise objects robustly even while the robot is moving, with an accuracy comparable to current techniques used on the iCub. Index Terms: spatial understanding, object localisation, humanoid robot, neural network, genetic programming
Keyphrases
towards spatial perception humanoid robot index term genetic programming kinematic model spatial understanding prior camera calibration position estimate current technique cm range accuracy localise object cartesian genetic programming trained artificial neural network object localisation neural network joint encoder position
