Goal-oriented behavior generation for visually-guided manipulation task

Sungmoon Jeong, Yunjung Park, Hiroaki Arie, Jun Tani, Minho Lee

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

1 Citation (Scopus)

Abstract

We propose a new neuro-robotics network architecture that can generate goal-oriented behavior for visually-guided multiple object manipulation task by a humanoid robot. For examples, given a "sequential hit" multiple objects task, the proposed network is able to modulate a humanoid robot's behavior by taking advantage of suitable timing for gazing, approaching and hitting the object and again for the other object. To solve a multiple object manipulation task via learning by examples, the current study considers two important mechanisms: (1) stereo visual attention with depth estimation for movement generation, dynamic neural networks for behavior generation and (2) their adaptive coordination. Stereo visual attention provides a goal-directed shift sequence in a visual scan path, and it can guide the generation of a behavior plan considering depth information for robot movement. The proposed model can simultaneously generate the corresponding sequences of goal-directed visual attention shifts and robot movement timing with regards to the current sensory states including visual stimuli and body postures. The experiments show that the proposed network can solve a multiple object manipulation task through learning, by which some novel behaviors without prior learning can be successfully generated.

Original languageEnglish
Title of host publicationLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Pages501-508
Number of pages8
Volume7062 LNCS
EditionPART 1
DOIs
Publication statusPublished - 2011
Externally publishedYes
Event18th International Conference on Neural Information Processing, ICONIP 2011 - Shanghai
Duration: 2011 Nov 132011 Nov 17

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
NumberPART 1
Volume7062 LNCS
ISSN (Print)03029743
ISSN (Electronic)16113349

Other

Other18th International Conference on Neural Information Processing, ICONIP 2011
CityShanghai
Period11/11/1311/11/17

Fingerprint

Manipulation
Robots
Visual Attention
Humanoid Robot
Timing
Robot
Network architecture
Dynamic Neural Networks
Depth Estimation
Robotics
Network Architecture
Neural networks
Hits
Object
Path
Experiments
Experiment
Movement
Learning
Vision

Keywords

  • behavior generation
  • Multiple object manipulation task
  • multiple time-scale recurrent neural networks
  • stereo visual attention
  • visual attention shifts

ASJC Scopus subject areas

  • Computer Science(all)
  • Theoretical Computer Science

Cite this

Jeong, S., Park, Y., Arie, H., Tani, J., & Lee, M. (2011). Goal-oriented behavior generation for visually-guided manipulation task. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (PART 1 ed., Vol. 7062 LNCS, pp. 501-508). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 7062 LNCS, No. PART 1). https://doi.org/10.1007/978-3-642-24955-6_60

Goal-oriented behavior generation for visually-guided manipulation task. / Jeong, Sungmoon; Park, Yunjung; Arie, Hiroaki; Tani, Jun; Lee, Minho.

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 7062 LNCS PART 1. ed. 2011. p. 501-508 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 7062 LNCS, No. PART 1).

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

Jeong, S, Park, Y, Arie, H, Tani, J & Lee, M 2011, Goal-oriented behavior generation for visually-guided manipulation task. in Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). PART 1 edn, vol. 7062 LNCS, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), no. PART 1, vol. 7062 LNCS, pp. 501-508, 18th International Conference on Neural Information Processing, ICONIP 2011, Shanghai, 11/11/13. https://doi.org/10.1007/978-3-642-24955-6_60
Jeong S, Park Y, Arie H, Tani J, Lee M. Goal-oriented behavior generation for visually-guided manipulation task. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). PART 1 ed. Vol. 7062 LNCS. 2011. p. 501-508. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); PART 1). https://doi.org/10.1007/978-3-642-24955-6_60
Jeong, Sungmoon ; Park, Yunjung ; Arie, Hiroaki ; Tani, Jun ; Lee, Minho. / Goal-oriented behavior generation for visually-guided manipulation task. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 7062 LNCS PART 1. ed. 2011. pp. 501-508 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); PART 1).
@inproceedings{288eb11bc7134a76b0a1326aa920e824,
title = "Goal-oriented behavior generation for visually-guided manipulation task",
abstract = "We propose a new neuro-robotics network architecture that can generate goal-oriented behavior for visually-guided multiple object manipulation task by a humanoid robot. For examples, given a {"}sequential hit{"} multiple objects task, the proposed network is able to modulate a humanoid robot's behavior by taking advantage of suitable timing for gazing, approaching and hitting the object and again for the other object. To solve a multiple object manipulation task via learning by examples, the current study considers two important mechanisms: (1) stereo visual attention with depth estimation for movement generation, dynamic neural networks for behavior generation and (2) their adaptive coordination. Stereo visual attention provides a goal-directed shift sequence in a visual scan path, and it can guide the generation of a behavior plan considering depth information for robot movement. The proposed model can simultaneously generate the corresponding sequences of goal-directed visual attention shifts and robot movement timing with regards to the current sensory states including visual stimuli and body postures. The experiments show that the proposed network can solve a multiple object manipulation task through learning, by which some novel behaviors without prior learning can be successfully generated.",
keywords = "behavior generation, Multiple object manipulation task, multiple time-scale recurrent neural networks, stereo visual attention, visual attention shifts",
author = "Sungmoon Jeong and Yunjung Park and Hiroaki Arie and Jun Tani and Minho Lee",
year = "2011",
doi = "10.1007/978-3-642-24955-6_60",
language = "English",
isbn = "9783642249549",
volume = "7062 LNCS",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
number = "PART 1",
pages = "501--508",
booktitle = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
edition = "PART 1",

}

TY - GEN

T1 - Goal-oriented behavior generation for visually-guided manipulation task

AU - Jeong, Sungmoon

AU - Park, Yunjung

AU - Arie, Hiroaki

AU - Tani, Jun

AU - Lee, Minho

PY - 2011

Y1 - 2011

N2 - We propose a new neuro-robotics network architecture that can generate goal-oriented behavior for visually-guided multiple object manipulation task by a humanoid robot. For examples, given a "sequential hit" multiple objects task, the proposed network is able to modulate a humanoid robot's behavior by taking advantage of suitable timing for gazing, approaching and hitting the object and again for the other object. To solve a multiple object manipulation task via learning by examples, the current study considers two important mechanisms: (1) stereo visual attention with depth estimation for movement generation, dynamic neural networks for behavior generation and (2) their adaptive coordination. Stereo visual attention provides a goal-directed shift sequence in a visual scan path, and it can guide the generation of a behavior plan considering depth information for robot movement. The proposed model can simultaneously generate the corresponding sequences of goal-directed visual attention shifts and robot movement timing with regards to the current sensory states including visual stimuli and body postures. The experiments show that the proposed network can solve a multiple object manipulation task through learning, by which some novel behaviors without prior learning can be successfully generated.

AB - We propose a new neuro-robotics network architecture that can generate goal-oriented behavior for visually-guided multiple object manipulation task by a humanoid robot. For examples, given a "sequential hit" multiple objects task, the proposed network is able to modulate a humanoid robot's behavior by taking advantage of suitable timing for gazing, approaching and hitting the object and again for the other object. To solve a multiple object manipulation task via learning by examples, the current study considers two important mechanisms: (1) stereo visual attention with depth estimation for movement generation, dynamic neural networks for behavior generation and (2) their adaptive coordination. Stereo visual attention provides a goal-directed shift sequence in a visual scan path, and it can guide the generation of a behavior plan considering depth information for robot movement. The proposed model can simultaneously generate the corresponding sequences of goal-directed visual attention shifts and robot movement timing with regards to the current sensory states including visual stimuli and body postures. The experiments show that the proposed network can solve a multiple object manipulation task through learning, by which some novel behaviors without prior learning can be successfully generated.

KW - behavior generation

KW - Multiple object manipulation task

KW - multiple time-scale recurrent neural networks

KW - stereo visual attention

KW - visual attention shifts

UR - http://www.scopus.com/inward/record.url?scp=81855190760&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=81855190760&partnerID=8YFLogxK

U2 - 10.1007/978-3-642-24955-6_60

DO - 10.1007/978-3-642-24955-6_60

M3 - Conference contribution

AN - SCOPUS:81855190760

SN - 9783642249549

VL - 7062 LNCS

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 501

EP - 508

BT - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

ER -