Dynamic motion generation by flexible-joint robot based on deep learning using images

Yuheng Wu, Kuniyuki Takahashi, Hiroki Yamada, Kitae Kim, Shingo Murata, Shigeki Sugano, Tetsuya Ogata

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

Abstract

Robots with flexible joints have recently been attracting attention from researchers because such robots can passively adapt to environmental changes and realize dynamic motion that uses inertia. In previous research, body-model acquisition using deep learning was proposed and dynamic motion learning was achieved. However, using the end-effector position as a visual feedback signal to train a robot limits what the robot can know to only the relation between the task and itself, instead of the relation between the environment and itself. In this research, we propose to use images as a feedback signal so that the robot can have a sense of the overall situation within the task environment. This motion learning is performed via deep learning using raw image data. In an experiment, we let a robot perform task motions once to acquire motor and image data. Then, we used a convolutional auto-encoder to extract image features from raw image data. The extracted image features were used in combination with motor data to train a recurrent neural network. As a result, motion learning through deep learning from image data allowed the robot to acquire environmental information and conduct tasks that require consideration of environmental changes, making use of its advantage of passive adaptation.

Original languageEnglish
Title of host publication2018 Joint IEEE 8th International Conference on Development and Learning and Epigenetic Robotics, ICDL-EpiRob 2018
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages169-174
Number of pages6
ISBN (Electronic)9781538661109
DOIs
Publication statusPublished - 2018 Sep 1
EventJoint 8th IEEE International Conference on Development and Learning and Epigenetic Robotics, ICDL-EpiRob 2018 - Tokyo, Japan
Duration: 2018 Sep 162018 Sep 20

Publication series

Name2018 Joint IEEE 8th International Conference on Development and Learning and Epigenetic Robotics, ICDL-EpiRob 2018

Conference

ConferenceJoint 8th IEEE International Conference on Development and Learning and Epigenetic Robotics, ICDL-EpiRob 2018
CountryJapan
CityTokyo
Period18/9/1618/9/20

Fingerprint

Joints
Robot
Learning
Robots
Motion
Sensory Feedback
Feedback
Recurrent neural networks
Research
Recurrent Neural Networks
End effectors
Encoder
Deep learning
Inertia
Research Personnel
Experiment
Experiments

ASJC Scopus subject areas

  • Computer Vision and Pattern Recognition
  • Control and Optimization
  • Behavioral Neuroscience
  • Developmental Neuroscience
  • Artificial Intelligence

Cite this

Wu, Y., Takahashi, K., Yamada, H., Kim, K., Murata, S., Sugano, S., & Ogata, T. (2018). Dynamic motion generation by flexible-joint robot based on deep learning using images. In 2018 Joint IEEE 8th International Conference on Development and Learning and Epigenetic Robotics, ICDL-EpiRob 2018 (pp. 169-174). [8761020] (2018 Joint IEEE 8th International Conference on Development and Learning and Epigenetic Robotics, ICDL-EpiRob 2018). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/DEVLRN.2018.8761020

Dynamic motion generation by flexible-joint robot based on deep learning using images. / Wu, Yuheng; Takahashi, Kuniyuki; Yamada, Hiroki; Kim, Kitae; Murata, Shingo; Sugano, Shigeki; Ogata, Tetsuya.

2018 Joint IEEE 8th International Conference on Development and Learning and Epigenetic Robotics, ICDL-EpiRob 2018. Institute of Electrical and Electronics Engineers Inc., 2018. p. 169-174 8761020 (2018 Joint IEEE 8th International Conference on Development and Learning and Epigenetic Robotics, ICDL-EpiRob 2018).

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

Wu, Y, Takahashi, K, Yamada, H, Kim, K, Murata, S, Sugano, S & Ogata, T 2018, Dynamic motion generation by flexible-joint robot based on deep learning using images. in 2018 Joint IEEE 8th International Conference on Development and Learning and Epigenetic Robotics, ICDL-EpiRob 2018., 8761020, 2018 Joint IEEE 8th International Conference on Development and Learning and Epigenetic Robotics, ICDL-EpiRob 2018, Institute of Electrical and Electronics Engineers Inc., pp. 169-174, Joint 8th IEEE International Conference on Development and Learning and Epigenetic Robotics, ICDL-EpiRob 2018, Tokyo, Japan, 18/9/16. https://doi.org/10.1109/DEVLRN.2018.8761020
Wu Y, Takahashi K, Yamada H, Kim K, Murata S, Sugano S et al. Dynamic motion generation by flexible-joint robot based on deep learning using images. In 2018 Joint IEEE 8th International Conference on Development and Learning and Epigenetic Robotics, ICDL-EpiRob 2018. Institute of Electrical and Electronics Engineers Inc. 2018. p. 169-174. 8761020. (2018 Joint IEEE 8th International Conference on Development and Learning and Epigenetic Robotics, ICDL-EpiRob 2018). https://doi.org/10.1109/DEVLRN.2018.8761020
Wu, Yuheng ; Takahashi, Kuniyuki ; Yamada, Hiroki ; Kim, Kitae ; Murata, Shingo ; Sugano, Shigeki ; Ogata, Tetsuya. / Dynamic motion generation by flexible-joint robot based on deep learning using images. 2018 Joint IEEE 8th International Conference on Development and Learning and Epigenetic Robotics, ICDL-EpiRob 2018. Institute of Electrical and Electronics Engineers Inc., 2018. pp. 169-174 (2018 Joint IEEE 8th International Conference on Development and Learning and Epigenetic Robotics, ICDL-EpiRob 2018).
@inproceedings{3f43ce63117e462d8277c0e02189a147,
title = "Dynamic motion generation by flexible-joint robot based on deep learning using images",
abstract = "Robots with flexible joints have recently been attracting attention from researchers because such robots can passively adapt to environmental changes and realize dynamic motion that uses inertia. In previous research, body-model acquisition using deep learning was proposed and dynamic motion learning was achieved. However, using the end-effector position as a visual feedback signal to train a robot limits what the robot can know to only the relation between the task and itself, instead of the relation between the environment and itself. In this research, we propose to use images as a feedback signal so that the robot can have a sense of the overall situation within the task environment. This motion learning is performed via deep learning using raw image data. In an experiment, we let a robot perform task motions once to acquire motor and image data. Then, we used a convolutional auto-encoder to extract image features from raw image data. The extracted image features were used in combination with motor data to train a recurrent neural network. As a result, motion learning through deep learning from image data allowed the robot to acquire environmental information and conduct tasks that require consideration of environmental changes, making use of its advantage of passive adaptation.",
author = "Yuheng Wu and Kuniyuki Takahashi and Hiroki Yamada and Kitae Kim and Shingo Murata and Shigeki Sugano and Tetsuya Ogata",
year = "2018",
month = "9",
day = "1",
doi = "10.1109/DEVLRN.2018.8761020",
language = "English",
series = "2018 Joint IEEE 8th International Conference on Development and Learning and Epigenetic Robotics, ICDL-EpiRob 2018",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
pages = "169--174",
booktitle = "2018 Joint IEEE 8th International Conference on Development and Learning and Epigenetic Robotics, ICDL-EpiRob 2018",

}

TY - GEN

T1 - Dynamic motion generation by flexible-joint robot based on deep learning using images

AU - Wu, Yuheng

AU - Takahashi, Kuniyuki

AU - Yamada, Hiroki

AU - Kim, Kitae

AU - Murata, Shingo

AU - Sugano, Shigeki

AU - Ogata, Tetsuya

PY - 2018/9/1

Y1 - 2018/9/1

N2 - Robots with flexible joints have recently been attracting attention from researchers because such robots can passively adapt to environmental changes and realize dynamic motion that uses inertia. In previous research, body-model acquisition using deep learning was proposed and dynamic motion learning was achieved. However, using the end-effector position as a visual feedback signal to train a robot limits what the robot can know to only the relation between the task and itself, instead of the relation between the environment and itself. In this research, we propose to use images as a feedback signal so that the robot can have a sense of the overall situation within the task environment. This motion learning is performed via deep learning using raw image data. In an experiment, we let a robot perform task motions once to acquire motor and image data. Then, we used a convolutional auto-encoder to extract image features from raw image data. The extracted image features were used in combination with motor data to train a recurrent neural network. As a result, motion learning through deep learning from image data allowed the robot to acquire environmental information and conduct tasks that require consideration of environmental changes, making use of its advantage of passive adaptation.

AB - Robots with flexible joints have recently been attracting attention from researchers because such robots can passively adapt to environmental changes and realize dynamic motion that uses inertia. In previous research, body-model acquisition using deep learning was proposed and dynamic motion learning was achieved. However, using the end-effector position as a visual feedback signal to train a robot limits what the robot can know to only the relation between the task and itself, instead of the relation between the environment and itself. In this research, we propose to use images as a feedback signal so that the robot can have a sense of the overall situation within the task environment. This motion learning is performed via deep learning using raw image data. In an experiment, we let a robot perform task motions once to acquire motor and image data. Then, we used a convolutional auto-encoder to extract image features from raw image data. The extracted image features were used in combination with motor data to train a recurrent neural network. As a result, motion learning through deep learning from image data allowed the robot to acquire environmental information and conduct tasks that require consideration of environmental changes, making use of its advantage of passive adaptation.

UR - http://www.scopus.com/inward/record.url?scp=85070380587&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85070380587&partnerID=8YFLogxK

U2 - 10.1109/DEVLRN.2018.8761020

DO - 10.1109/DEVLRN.2018.8761020

M3 - Conference contribution

T3 - 2018 Joint IEEE 8th International Conference on Development and Learning and Epigenetic Robotics, ICDL-EpiRob 2018

SP - 169

EP - 174

BT - 2018 Joint IEEE 8th International Conference on Development and Learning and Epigenetic Robotics, ICDL-EpiRob 2018

PB - Institute of Electrical and Electronics Engineers Inc.

ER -