Versatile In-Hand Manipulation of Objects with Different Sizes and Shapes Using Neural Networks

Satoshi Funabashi, Alexander Schmitz, Takashi Sato, Sophon Somlor, Shigeki Sugano

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

Abstract

Changing the grasping posture of objects within a robot hand is hard to achieve, especially if the objects are of various shape and size. In this paper we use a neural network to learn such manipulation with variously sized and shaped objects. The TWENDY-ONE hand possesses various properties that are effective for in-hand manipulation: a high number of actuated joints, passive degrees of freedom and soft skin, six-axis force/torque (F /T) sensors in each fingertip and distributed tactile sensors in the soft skin. The object size information is extracted from the initial grasping posture. The training data includes tactile and the object information. After training the neural network, the robot is able to manipulate objects of not only trained but also untrained size and shape. The results show the importance of size and tactile information. Importantly, the features extracted by a stacked autoencoder (trained with a larger dataset) could reduce the number of required training samples for supervised learning of in-hand manipulation.

Original languageEnglish
Title of host publication2018 IEEE-RAS 18th International Conference on Humanoid Robots, Humanoids 2018
PublisherIEEE Computer Society
Pages768-775
Number of pages8
ISBN (Electronic)9781538672839
DOIs
Publication statusPublished - 2019 Jan 23
Event18th IEEE-RAS International Conference on Humanoid Robots, Humanoids 2018 - Beijing, China
Duration: 2018 Nov 62018 Nov 9

Publication series

NameIEEE-RAS International Conference on Humanoid Robots
Volume2018-November
ISSN (Print)2164-0572
ISSN (Electronic)2164-0580

Conference

Conference18th IEEE-RAS International Conference on Humanoid Robots, Humanoids 2018
CountryChina
CityBeijing
Period18/11/618/11/9

Fingerprint

Skin
Robots
Neural networks
Supervised learning
Sensors
End effectors
Torque

ASJC Scopus subject areas

  • Artificial Intelligence
  • Computer Vision and Pattern Recognition
  • Hardware and Architecture
  • Human-Computer Interaction
  • Electrical and Electronic Engineering

Cite this

Funabashi, S., Schmitz, A., Sato, T., Somlor, S., & Sugano, S. (2019). Versatile In-Hand Manipulation of Objects with Different Sizes and Shapes Using Neural Networks. In 2018 IEEE-RAS 18th International Conference on Humanoid Robots, Humanoids 2018 (pp. 768-775). [8624961] (IEEE-RAS International Conference on Humanoid Robots; Vol. 2018-November). IEEE Computer Society. https://doi.org/10.1109/HUMANOIDS.2018.8624961

Versatile In-Hand Manipulation of Objects with Different Sizes and Shapes Using Neural Networks. / Funabashi, Satoshi; Schmitz, Alexander; Sato, Takashi; Somlor, Sophon; Sugano, Shigeki.

2018 IEEE-RAS 18th International Conference on Humanoid Robots, Humanoids 2018. IEEE Computer Society, 2019. p. 768-775 8624961 (IEEE-RAS International Conference on Humanoid Robots; Vol. 2018-November).

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

Funabashi, S, Schmitz, A, Sato, T, Somlor, S & Sugano, S 2019, Versatile In-Hand Manipulation of Objects with Different Sizes and Shapes Using Neural Networks. in 2018 IEEE-RAS 18th International Conference on Humanoid Robots, Humanoids 2018., 8624961, IEEE-RAS International Conference on Humanoid Robots, vol. 2018-November, IEEE Computer Society, pp. 768-775, 18th IEEE-RAS International Conference on Humanoid Robots, Humanoids 2018, Beijing, China, 18/11/6. https://doi.org/10.1109/HUMANOIDS.2018.8624961
Funabashi S, Schmitz A, Sato T, Somlor S, Sugano S. Versatile In-Hand Manipulation of Objects with Different Sizes and Shapes Using Neural Networks. In 2018 IEEE-RAS 18th International Conference on Humanoid Robots, Humanoids 2018. IEEE Computer Society. 2019. p. 768-775. 8624961. (IEEE-RAS International Conference on Humanoid Robots). https://doi.org/10.1109/HUMANOIDS.2018.8624961
Funabashi, Satoshi ; Schmitz, Alexander ; Sato, Takashi ; Somlor, Sophon ; Sugano, Shigeki. / Versatile In-Hand Manipulation of Objects with Different Sizes and Shapes Using Neural Networks. 2018 IEEE-RAS 18th International Conference on Humanoid Robots, Humanoids 2018. IEEE Computer Society, 2019. pp. 768-775 (IEEE-RAS International Conference on Humanoid Robots).
@inproceedings{8f9fca8fcc1346bb9fc072c01b8c8db3,
title = "Versatile In-Hand Manipulation of Objects with Different Sizes and Shapes Using Neural Networks",
abstract = "Changing the grasping posture of objects within a robot hand is hard to achieve, especially if the objects are of various shape and size. In this paper we use a neural network to learn such manipulation with variously sized and shaped objects. The TWENDY-ONE hand possesses various properties that are effective for in-hand manipulation: a high number of actuated joints, passive degrees of freedom and soft skin, six-axis force/torque (F /T) sensors in each fingertip and distributed tactile sensors in the soft skin. The object size information is extracted from the initial grasping posture. The training data includes tactile and the object information. After training the neural network, the robot is able to manipulate objects of not only trained but also untrained size and shape. The results show the importance of size and tactile information. Importantly, the features extracted by a stacked autoencoder (trained with a larger dataset) could reduce the number of required training samples for supervised learning of in-hand manipulation.",
author = "Satoshi Funabashi and Alexander Schmitz and Takashi Sato and Sophon Somlor and Shigeki Sugano",
year = "2019",
month = "1",
day = "23",
doi = "10.1109/HUMANOIDS.2018.8624961",
language = "English",
series = "IEEE-RAS International Conference on Humanoid Robots",
publisher = "IEEE Computer Society",
pages = "768--775",
booktitle = "2018 IEEE-RAS 18th International Conference on Humanoid Robots, Humanoids 2018",

}

TY - GEN

T1 - Versatile In-Hand Manipulation of Objects with Different Sizes and Shapes Using Neural Networks

AU - Funabashi, Satoshi

AU - Schmitz, Alexander

AU - Sato, Takashi

AU - Somlor, Sophon

AU - Sugano, Shigeki

PY - 2019/1/23

Y1 - 2019/1/23

N2 - Changing the grasping posture of objects within a robot hand is hard to achieve, especially if the objects are of various shape and size. In this paper we use a neural network to learn such manipulation with variously sized and shaped objects. The TWENDY-ONE hand possesses various properties that are effective for in-hand manipulation: a high number of actuated joints, passive degrees of freedom and soft skin, six-axis force/torque (F /T) sensors in each fingertip and distributed tactile sensors in the soft skin. The object size information is extracted from the initial grasping posture. The training data includes tactile and the object information. After training the neural network, the robot is able to manipulate objects of not only trained but also untrained size and shape. The results show the importance of size and tactile information. Importantly, the features extracted by a stacked autoencoder (trained with a larger dataset) could reduce the number of required training samples for supervised learning of in-hand manipulation.

AB - Changing the grasping posture of objects within a robot hand is hard to achieve, especially if the objects are of various shape and size. In this paper we use a neural network to learn such manipulation with variously sized and shaped objects. The TWENDY-ONE hand possesses various properties that are effective for in-hand manipulation: a high number of actuated joints, passive degrees of freedom and soft skin, six-axis force/torque (F /T) sensors in each fingertip and distributed tactile sensors in the soft skin. The object size information is extracted from the initial grasping posture. The training data includes tactile and the object information. After training the neural network, the robot is able to manipulate objects of not only trained but also untrained size and shape. The results show the importance of size and tactile information. Importantly, the features extracted by a stacked autoencoder (trained with a larger dataset) could reduce the number of required training samples for supervised learning of in-hand manipulation.

UR - http://www.scopus.com/inward/record.url?scp=85062258028&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85062258028&partnerID=8YFLogxK

U2 - 10.1109/HUMANOIDS.2018.8624961

DO - 10.1109/HUMANOIDS.2018.8624961

M3 - Conference contribution

AN - SCOPUS:85062258028

T3 - IEEE-RAS International Conference on Humanoid Robots

SP - 768

EP - 775

BT - 2018 IEEE-RAS 18th International Conference on Humanoid Robots, Humanoids 2018

PB - IEEE Computer Society

ER -