SKID RAW: Skill Discovery from Raw Trajectories

Research output: Contribution to journalArticleResearchpeer-review

Standard

SKID RAW: Skill Discovery from Raw Trajectories. / Tanneberg, Daniel; Ploeger, Kai; Rueckert, Elmar et al.
In: IEEE robotics and automation letters, Vol. 6, No. 3, 9387162, 07.2021, p. 4696-4703.

Research output: Contribution to journalArticleResearchpeer-review

Harvard

Vancouver

Tanneberg D, Ploeger K, Rueckert E, Peters J. SKID RAW: Skill Discovery from Raw Trajectories. IEEE robotics and automation letters. 2021 Jul;6(3):4696-4703. 9387162. Epub 2021 Mar 25. doi: 10.1109/LRA.2021.3068891

Author

Tanneberg, Daniel ; Ploeger, Kai ; Rueckert, Elmar et al. / SKID RAW : Skill Discovery from Raw Trajectories. In: IEEE robotics and automation letters. 2021 ; Vol. 6, No. 3. pp. 4696-4703.

Bibtex - Download

@article{2ed96ea757d8409292bc2d5f243d6dfe,
title = "SKID RAW: Skill Discovery from Raw Trajectories",
abstract = "Integrating robots in complex everyday environments requires a multitude of problems to be solved. One crucial feature among those is to equip robots with a mechanism for teaching them a new task in an easy and natural way. When teaching tasks that involve sequences of different skills, with varying order and number of these skills, it is desirable to only demonstrate full task executions instead of all individual skills. For this purpose, we propose a novel approach that simultaneously learns to segment trajectories into reoccurring patterns and the skills to reconstruct these patterns from unlabelled demonstrations without further supervision. Moreover, the approach learns a skill conditioning that can be used to understand possible sequences of skills, a practical mechanism to be used in, for example, human-robot-interactions for a more intelligent and adaptive robot behaviour. The Bayesian and variational inference based approach is evaluated on synthetic and real human demonstrations with varying complexities and dimensionality, showing the successful learning of segmentations and skill libraries from unlabelled data.",
keywords = "Deep learning methods, learning categories and concepts, representation learning, robot motion control, learning, movement primitives",
author = "Daniel Tanneberg and Kai Ploeger and Elmar Rueckert and Jan Peters",
note = "Publisher Copyright: {\textcopyright} 2016 IEEE.",
year = "2021",
month = jul,
doi = "10.1109/LRA.2021.3068891",
language = "English",
volume = "6",
pages = "4696--4703",
journal = " IEEE robotics and automation letters",
issn = "2377-3766",
publisher = "Institute of Electrical and Electronics Engineers",
number = "3",

}

RIS (suitable for import to EndNote) - Download

TY - JOUR

T1 - SKID RAW

T2 - Skill Discovery from Raw Trajectories

AU - Tanneberg, Daniel

AU - Ploeger, Kai

AU - Rueckert, Elmar

AU - Peters, Jan

N1 - Publisher Copyright: © 2016 IEEE.

PY - 2021/7

Y1 - 2021/7

N2 - Integrating robots in complex everyday environments requires a multitude of problems to be solved. One crucial feature among those is to equip robots with a mechanism for teaching them a new task in an easy and natural way. When teaching tasks that involve sequences of different skills, with varying order and number of these skills, it is desirable to only demonstrate full task executions instead of all individual skills. For this purpose, we propose a novel approach that simultaneously learns to segment trajectories into reoccurring patterns and the skills to reconstruct these patterns from unlabelled demonstrations without further supervision. Moreover, the approach learns a skill conditioning that can be used to understand possible sequences of skills, a practical mechanism to be used in, for example, human-robot-interactions for a more intelligent and adaptive robot behaviour. The Bayesian and variational inference based approach is evaluated on synthetic and real human demonstrations with varying complexities and dimensionality, showing the successful learning of segmentations and skill libraries from unlabelled data.

AB - Integrating robots in complex everyday environments requires a multitude of problems to be solved. One crucial feature among those is to equip robots with a mechanism for teaching them a new task in an easy and natural way. When teaching tasks that involve sequences of different skills, with varying order and number of these skills, it is desirable to only demonstrate full task executions instead of all individual skills. For this purpose, we propose a novel approach that simultaneously learns to segment trajectories into reoccurring patterns and the skills to reconstruct these patterns from unlabelled demonstrations without further supervision. Moreover, the approach learns a skill conditioning that can be used to understand possible sequences of skills, a practical mechanism to be used in, for example, human-robot-interactions for a more intelligent and adaptive robot behaviour. The Bayesian and variational inference based approach is evaluated on synthetic and real human demonstrations with varying complexities and dimensionality, showing the successful learning of segmentations and skill libraries from unlabelled data.

KW - Deep learning methods

KW - learning categories and concepts

KW - representation learning

KW - robot motion control

KW - learning

KW - movement primitives

UR - http://www.scopus.com/inward/record.url?scp=85103301482&partnerID=8YFLogxK

UR - https://cps.unileoben.ac.at/wp/RAL2021Tanneberg.pdf

U2 - 10.1109/LRA.2021.3068891

DO - 10.1109/LRA.2021.3068891

M3 - Article

AN - SCOPUS:85103301482

VL - 6

SP - 4696

EP - 4703

JO - IEEE robotics and automation letters

JF - IEEE robotics and automation letters

SN - 2377-3766

IS - 3

M1 - 9387162

ER -