Kinesthetic bootstrapping: Teaching motor skills to humanoid robots through physical interaction

Hani Ben Amor, Erik Berger, David Vogt, Bernhard Jung

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

31 Citations (Scopus)

Abstract

Programming of complex motor skills for humanoid robots can be a time intensive task, particularly within conventional textual or GUI-driven programming paradigms. Addressing this drawback, we propose a new programming-by-demonstration method called Kinesthetic Bootstrapping for teaching motor skills to humanoid robots by means of intuitive physical interactions. Here, "programming" simply consists of manually moving the robot's joints so as to demonstrate the skill in mind. The bootstrapping algorithm then generates a low-dimensional model of the demonstrated postures. To find a trajectory through this posture space that corresponds to a robust robot motion, a learning phase takes place in a physics-based virtual environment. The virtual robot's motion is optimized via a genetic algorithm and the result is transferred back to the physical robot. The method has been successfully applied to the learning of various complex motor skills such as walking and standing up.

Original languageEnglish (US)
Title of host publicationLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Pages492-499
Number of pages8
Volume5803 LNAI
DOIs
StatePublished - 2009
Externally publishedYes
Event32nd Annual German Conference on Artificial Intelligence, KI 2009 - Paderborn, Germany
Duration: Sep 15 2009Sep 18 2009

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume5803 LNAI
ISSN (Print)03029743
ISSN (Electronic)16113349

Other

Other32nd Annual German Conference on Artificial Intelligence, KI 2009
CountryGermany
CityPaderborn
Period9/15/099/18/09

Fingerprint

Humanoid Robot
Bootstrapping
Teaching
Programming
Robot
Robots
Interaction
Motion
Virtual Environments
Intuitive
Paradigm
Physics
Genetic Algorithm
Trajectory
Graphical user interfaces
Virtual reality
Skills
Demonstrations
Genetic algorithms
Trajectories

ASJC Scopus subject areas

  • Computer Science(all)
  • Theoretical Computer Science

Cite this

Ben Amor, H., Berger, E., Vogt, D., & Jung, B. (2009). Kinesthetic bootstrapping: Teaching motor skills to humanoid robots through physical interaction. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 5803 LNAI, pp. 492-499). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 5803 LNAI). https://doi.org/10.1007/978-3-642-04617-9_62

Kinesthetic bootstrapping : Teaching motor skills to humanoid robots through physical interaction. / Ben Amor, Hani; Berger, Erik; Vogt, David; Jung, Bernhard.

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 5803 LNAI 2009. p. 492-499 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 5803 LNAI).

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

Ben Amor, H, Berger, E, Vogt, D & Jung, B 2009, Kinesthetic bootstrapping: Teaching motor skills to humanoid robots through physical interaction. in Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). vol. 5803 LNAI, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 5803 LNAI, pp. 492-499, 32nd Annual German Conference on Artificial Intelligence, KI 2009, Paderborn, Germany, 9/15/09. https://doi.org/10.1007/978-3-642-04617-9_62
Ben Amor H, Berger E, Vogt D, Jung B. Kinesthetic bootstrapping: Teaching motor skills to humanoid robots through physical interaction. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 5803 LNAI. 2009. p. 492-499. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-642-04617-9_62
Ben Amor, Hani ; Berger, Erik ; Vogt, David ; Jung, Bernhard. / Kinesthetic bootstrapping : Teaching motor skills to humanoid robots through physical interaction. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 5803 LNAI 2009. pp. 492-499 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{de1422366c1b4d059033849779965c88,
title = "Kinesthetic bootstrapping: Teaching motor skills to humanoid robots through physical interaction",
abstract = "Programming of complex motor skills for humanoid robots can be a time intensive task, particularly within conventional textual or GUI-driven programming paradigms. Addressing this drawback, we propose a new programming-by-demonstration method called Kinesthetic Bootstrapping for teaching motor skills to humanoid robots by means of intuitive physical interactions. Here, {"}programming{"} simply consists of manually moving the robot's joints so as to demonstrate the skill in mind. The bootstrapping algorithm then generates a low-dimensional model of the demonstrated postures. To find a trajectory through this posture space that corresponds to a robust robot motion, a learning phase takes place in a physics-based virtual environment. The virtual robot's motion is optimized via a genetic algorithm and the result is transferred back to the physical robot. The method has been successfully applied to the learning of various complex motor skills such as walking and standing up.",
author = "{Ben Amor}, Hani and Erik Berger and David Vogt and Bernhard Jung",
year = "2009",
doi = "10.1007/978-3-642-04617-9_62",
language = "English (US)",
isbn = "3642046169",
volume = "5803 LNAI",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
pages = "492--499",
booktitle = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

}

TY - GEN

T1 - Kinesthetic bootstrapping

T2 - Teaching motor skills to humanoid robots through physical interaction

AU - Ben Amor, Hani

AU - Berger, Erik

AU - Vogt, David

AU - Jung, Bernhard

PY - 2009

Y1 - 2009

N2 - Programming of complex motor skills for humanoid robots can be a time intensive task, particularly within conventional textual or GUI-driven programming paradigms. Addressing this drawback, we propose a new programming-by-demonstration method called Kinesthetic Bootstrapping for teaching motor skills to humanoid robots by means of intuitive physical interactions. Here, "programming" simply consists of manually moving the robot's joints so as to demonstrate the skill in mind. The bootstrapping algorithm then generates a low-dimensional model of the demonstrated postures. To find a trajectory through this posture space that corresponds to a robust robot motion, a learning phase takes place in a physics-based virtual environment. The virtual robot's motion is optimized via a genetic algorithm and the result is transferred back to the physical robot. The method has been successfully applied to the learning of various complex motor skills such as walking and standing up.

AB - Programming of complex motor skills for humanoid robots can be a time intensive task, particularly within conventional textual or GUI-driven programming paradigms. Addressing this drawback, we propose a new programming-by-demonstration method called Kinesthetic Bootstrapping for teaching motor skills to humanoid robots by means of intuitive physical interactions. Here, "programming" simply consists of manually moving the robot's joints so as to demonstrate the skill in mind. The bootstrapping algorithm then generates a low-dimensional model of the demonstrated postures. To find a trajectory through this posture space that corresponds to a robust robot motion, a learning phase takes place in a physics-based virtual environment. The virtual robot's motion is optimized via a genetic algorithm and the result is transferred back to the physical robot. The method has been successfully applied to the learning of various complex motor skills such as walking and standing up.

UR - http://www.scopus.com/inward/record.url?scp=76649103008&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=76649103008&partnerID=8YFLogxK

U2 - 10.1007/978-3-642-04617-9_62

DO - 10.1007/978-3-642-04617-9_62

M3 - Conference contribution

SN - 3642046169

SN - 9783642046162

VL - 5803 LNAI

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

SP - 492

EP - 499

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

ER -