Learning to grasp objects based on ensemble learning combining simulation data and real data

Yong Ho Na, Hyunjun Jo, Jae-Bok Song

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

In this study, deep learning based grasping using a robot has been discussed. A large amount of training data is required for good performance in deep learning. The training data is usually collected with a real robot. However, it is difficult to collect the data sufficient for training the network in terms of time and cost. Therefore, this study presents a method for collecting the training data based on a robot simulator as well as a real robot. The simulation system is composed of a robot, the work environment, and a 2-finger gripper. The convolutional neural network (CNN) was used for training where its input is the RGB image of the object and its output is the pose of the gripper. Furthermore, the ensemble learning method was used to combine real data and simulation data. It is shown that the ensemble learning method that combines multiple classifiers can lead to a higher grasping success rate than a single classifier.

Original languageEnglish
Title of host publicationICCAS 2017 - 2017 17th International Conference on Control, Automation and Systems - Proceedings
PublisherIEEE Computer Society
Pages1030-1034
Number of pages5
Volume2017-October
ISBN (Electronic)9788993215137
DOIs
Publication statusPublished - 2017 Dec 13
Event17th International Conference on Control, Automation and Systems, ICCAS 2017 - Jeju, Korea, Republic of
Duration: 2017 Oct 182017 Oct 21

Other

Other17th International Conference on Control, Automation and Systems, ICCAS 2017
CountryKorea, Republic of
CityJeju
Period17/10/1817/10/21

Fingerprint

Robots
Grippers
Classifiers
Simulators
Neural networks
Costs
Deep learning

Keywords

  • Convolutional Neural Network
  • Deep learning
  • Ensemble learning
  • Robot grasping
  • Simulator robot

ASJC Scopus subject areas

  • Artificial Intelligence
  • Computer Science Applications
  • Control and Systems Engineering
  • Electrical and Electronic Engineering

Cite this

Na, Y. H., Jo, H., & Song, J-B. (2017). Learning to grasp objects based on ensemble learning combining simulation data and real data. In ICCAS 2017 - 2017 17th International Conference on Control, Automation and Systems - Proceedings (Vol. 2017-October, pp. 1030-1034). IEEE Computer Society. https://doi.org/10.23919/ICCAS.2017.8204368

Learning to grasp objects based on ensemble learning combining simulation data and real data. / Na, Yong Ho; Jo, Hyunjun; Song, Jae-Bok.

ICCAS 2017 - 2017 17th International Conference on Control, Automation and Systems - Proceedings. Vol. 2017-October IEEE Computer Society, 2017. p. 1030-1034.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Na, YH, Jo, H & Song, J-B 2017, Learning to grasp objects based on ensemble learning combining simulation data and real data. in ICCAS 2017 - 2017 17th International Conference on Control, Automation and Systems - Proceedings. vol. 2017-October, IEEE Computer Society, pp. 1030-1034, 17th International Conference on Control, Automation and Systems, ICCAS 2017, Jeju, Korea, Republic of, 17/10/18. https://doi.org/10.23919/ICCAS.2017.8204368
Na YH, Jo H, Song J-B. Learning to grasp objects based on ensemble learning combining simulation data and real data. In ICCAS 2017 - 2017 17th International Conference on Control, Automation and Systems - Proceedings. Vol. 2017-October. IEEE Computer Society. 2017. p. 1030-1034 https://doi.org/10.23919/ICCAS.2017.8204368
Na, Yong Ho ; Jo, Hyunjun ; Song, Jae-Bok. / Learning to grasp objects based on ensemble learning combining simulation data and real data. ICCAS 2017 - 2017 17th International Conference on Control, Automation and Systems - Proceedings. Vol. 2017-October IEEE Computer Society, 2017. pp. 1030-1034
@inproceedings{f2811c04a6f44c89ab67eadc442e050d,
title = "Learning to grasp objects based on ensemble learning combining simulation data and real data",
abstract = "In this study, deep learning based grasping using a robot has been discussed. A large amount of training data is required for good performance in deep learning. The training data is usually collected with a real robot. However, it is difficult to collect the data sufficient for training the network in terms of time and cost. Therefore, this study presents a method for collecting the training data based on a robot simulator as well as a real robot. The simulation system is composed of a robot, the work environment, and a 2-finger gripper. The convolutional neural network (CNN) was used for training where its input is the RGB image of the object and its output is the pose of the gripper. Furthermore, the ensemble learning method was used to combine real data and simulation data. It is shown that the ensemble learning method that combines multiple classifiers can lead to a higher grasping success rate than a single classifier.",
keywords = "Convolutional Neural Network, Deep learning, Ensemble learning, Robot grasping, Simulator robot",
author = "Na, {Yong Ho} and Hyunjun Jo and Jae-Bok Song",
year = "2017",
month = "12",
day = "13",
doi = "10.23919/ICCAS.2017.8204368",
language = "English",
volume = "2017-October",
pages = "1030--1034",
booktitle = "ICCAS 2017 - 2017 17th International Conference on Control, Automation and Systems - Proceedings",
publisher = "IEEE Computer Society",

}

TY - GEN

T1 - Learning to grasp objects based on ensemble learning combining simulation data and real data

AU - Na, Yong Ho

AU - Jo, Hyunjun

AU - Song, Jae-Bok

PY - 2017/12/13

Y1 - 2017/12/13

N2 - In this study, deep learning based grasping using a robot has been discussed. A large amount of training data is required for good performance in deep learning. The training data is usually collected with a real robot. However, it is difficult to collect the data sufficient for training the network in terms of time and cost. Therefore, this study presents a method for collecting the training data based on a robot simulator as well as a real robot. The simulation system is composed of a robot, the work environment, and a 2-finger gripper. The convolutional neural network (CNN) was used for training where its input is the RGB image of the object and its output is the pose of the gripper. Furthermore, the ensemble learning method was used to combine real data and simulation data. It is shown that the ensemble learning method that combines multiple classifiers can lead to a higher grasping success rate than a single classifier.

AB - In this study, deep learning based grasping using a robot has been discussed. A large amount of training data is required for good performance in deep learning. The training data is usually collected with a real robot. However, it is difficult to collect the data sufficient for training the network in terms of time and cost. Therefore, this study presents a method for collecting the training data based on a robot simulator as well as a real robot. The simulation system is composed of a robot, the work environment, and a 2-finger gripper. The convolutional neural network (CNN) was used for training where its input is the RGB image of the object and its output is the pose of the gripper. Furthermore, the ensemble learning method was used to combine real data and simulation data. It is shown that the ensemble learning method that combines multiple classifiers can lead to a higher grasping success rate than a single classifier.

KW - Convolutional Neural Network

KW - Deep learning

KW - Ensemble learning

KW - Robot grasping

KW - Simulator robot

UR - http://www.scopus.com/inward/record.url?scp=85044455650&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85044455650&partnerID=8YFLogxK

U2 - 10.23919/ICCAS.2017.8204368

DO - 10.23919/ICCAS.2017.8204368

M3 - Conference contribution

AN - SCOPUS:85044455650

VL - 2017-October

SP - 1030

EP - 1034

BT - ICCAS 2017 - 2017 17th International Conference on Control, Automation and Systems - Proceedings

PB - IEEE Computer Society

ER -