Online Learning of the Inverse Dynamics with Parallel Drifting Gaussian Processes: Implementation of an Approach for Feedforward Control of a Parallel Kinematic Industrial Robot

Publikation: Beitrag in Buch/Bericht/Sammelwerk/KonferenzbandAufsatz in KonferenzbandForschungPeer-Review

Autoren

Organisationseinheiten

Forschungs-netzwerk anzeigen

Details

OriginalspracheEnglisch
Titel des SammelwerksProceedings of 2019 IEEE International Conference on Mechatronics and Automation, ICMA 2019
Herausgeber (Verlag)Institute of Electrical and Electronics Engineers Inc.
Seiten962-969
Seitenumfang8
ISBN (elektronisch)978-1-7281-1699-0
ISBN (Print)978-1-7281-1698-3, 978-1-7281-1700-3
PublikationsstatusVeröffentlicht - Aug. 2019
Veranstaltung16th IEEE International Conference on Mechatronics and Automation, ICMA 2019 - Tianjin, China
Dauer: 4 Aug. 20197 Aug. 2019

Publikationsreihe

NameProceedings of 2019 IEEE International Conference on Mechatronics and Automation, ICMA 2019
ISSN (Print)2152-7431
ISSN (elektronisch)2152-744X

Abstract

The present paper deals with an online approach to learn the inverse dynamics of any robot. This is realized by the use of Gaussian Processes drifting parallel along the system data. An extension by a database enables the efficient use of data points from the past. The central component of this work is the implementation of such a method in a controller in order to achieve the actual goal: the feedforward control of an industrial robot by means of machine learning. This is done by splitting the procedure into two threads running parallel so that the prediction is decoupled from the computing-intensive training of the models. Experiments show that the method reduces the tracking errors more clearly than an elaborately identified rigid body model including friction. For a defined trajectory, the squared areas of the tracking errors of all axes are reduced by more than 54% compared to motion without pre-control. In addition, a highly dynamic pick-and-place experiment is used to investigate the possible changes in system dynamics. Compared to an offline trained model, the approximation error of the proposed online approach is smaller for the remaining time of the experiment after an initial phase. Furthermore, this error is smaller throughout the experiment for online learning with parallel drifting Gaussian Processes than when using a single one.

ASJC Scopus Sachgebiete

Zitieren

Online Learning of the Inverse Dynamics with Parallel Drifting Gaussian Processes: Implementation of an Approach for Feedforward Control of a Parallel Kinematic Industrial Robot. / Habich, Tim-Lukas; Kaczor, Daniel; Tappe, Svenja et al.
Proceedings of 2019 IEEE International Conference on Mechatronics and Automation, ICMA 2019. Institute of Electrical and Electronics Engineers Inc., 2019. S. 962-969 8816298 (Proceedings of 2019 IEEE International Conference on Mechatronics and Automation, ICMA 2019).

Publikation: Beitrag in Buch/Bericht/Sammelwerk/KonferenzbandAufsatz in KonferenzbandForschungPeer-Review

Habich, T-L, Kaczor, D, Tappe, S & Ortmaier, T 2019, Online Learning of the Inverse Dynamics with Parallel Drifting Gaussian Processes: Implementation of an Approach for Feedforward Control of a Parallel Kinematic Industrial Robot. in Proceedings of 2019 IEEE International Conference on Mechatronics and Automation, ICMA 2019., 8816298, Proceedings of 2019 IEEE International Conference on Mechatronics and Automation, ICMA 2019, Institute of Electrical and Electronics Engineers Inc., S. 962-969, 16th IEEE International Conference on Mechatronics and Automation, ICMA 2019, Tianjin, China, 4 Aug. 2019. https://doi.org/10.1109/ICMA.2019.8816298
Habich, T.-L., Kaczor, D., Tappe, S., & Ortmaier, T. (2019). Online Learning of the Inverse Dynamics with Parallel Drifting Gaussian Processes: Implementation of an Approach for Feedforward Control of a Parallel Kinematic Industrial Robot. In Proceedings of 2019 IEEE International Conference on Mechatronics and Automation, ICMA 2019 (S. 962-969). Artikel 8816298 (Proceedings of 2019 IEEE International Conference on Mechatronics and Automation, ICMA 2019). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/ICMA.2019.8816298
Habich TL, Kaczor D, Tappe S, Ortmaier T. Online Learning of the Inverse Dynamics with Parallel Drifting Gaussian Processes: Implementation of an Approach for Feedforward Control of a Parallel Kinematic Industrial Robot. in Proceedings of 2019 IEEE International Conference on Mechatronics and Automation, ICMA 2019. Institute of Electrical and Electronics Engineers Inc. 2019. S. 962-969. 8816298. (Proceedings of 2019 IEEE International Conference on Mechatronics and Automation, ICMA 2019). doi: 10.1109/ICMA.2019.8816298
Habich, Tim-Lukas ; Kaczor, Daniel ; Tappe, Svenja et al. / Online Learning of the Inverse Dynamics with Parallel Drifting Gaussian Processes : Implementation of an Approach for Feedforward Control of a Parallel Kinematic Industrial Robot. Proceedings of 2019 IEEE International Conference on Mechatronics and Automation, ICMA 2019. Institute of Electrical and Electronics Engineers Inc., 2019. S. 962-969 (Proceedings of 2019 IEEE International Conference on Mechatronics and Automation, ICMA 2019).
Download
@inproceedings{e14610b3d8ca4a8a91be646745430c94,
title = "Online Learning of the Inverse Dynamics with Parallel Drifting Gaussian Processes: Implementation of an Approach for Feedforward Control of a Parallel Kinematic Industrial Robot",
abstract = "The present paper deals with an online approach to learn the inverse dynamics of any robot. This is realized by the use of Gaussian Processes drifting parallel along the system data. An extension by a database enables the efficient use of data points from the past. The central component of this work is the implementation of such a method in a controller in order to achieve the actual goal: the feedforward control of an industrial robot by means of machine learning. This is done by splitting the procedure into two threads running parallel so that the prediction is decoupled from the computing-intensive training of the models. Experiments show that the method reduces the tracking errors more clearly than an elaborately identified rigid body model including friction. For a defined trajectory, the squared areas of the tracking errors of all axes are reduced by more than 54% compared to motion without pre-control. In addition, a highly dynamic pick-and-place experiment is used to investigate the possible changes in system dynamics. Compared to an offline trained model, the approximation error of the proposed online approach is smaller for the remaining time of the experiment after an initial phase. Furthermore, this error is smaller throughout the experiment for online learning with parallel drifting Gaussian Processes than when using a single one.",
keywords = "Feedforward control, Gaussian Process, Implementation, Inverse dynamics, Online learning",
author = "Tim-Lukas Habich and Daniel Kaczor and Svenja Tappe and Tobias Ortmaier",
year = "2019",
month = aug,
doi = "10.1109/ICMA.2019.8816298",
language = "English",
isbn = "978-1-7281-1698-3",
series = "Proceedings of 2019 IEEE International Conference on Mechatronics and Automation, ICMA 2019",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
pages = "962--969",
booktitle = "Proceedings of 2019 IEEE International Conference on Mechatronics and Automation, ICMA 2019",
address = "United States",
note = "16th IEEE International Conference on Mechatronics and Automation, ICMA 2019 ; Conference date: 04-08-2019 Through 07-08-2019",

}

Download

TY - GEN

T1 - Online Learning of the Inverse Dynamics with Parallel Drifting Gaussian Processes

T2 - 16th IEEE International Conference on Mechatronics and Automation, ICMA 2019

AU - Habich, Tim-Lukas

AU - Kaczor, Daniel

AU - Tappe, Svenja

AU - Ortmaier, Tobias

PY - 2019/8

Y1 - 2019/8

N2 - The present paper deals with an online approach to learn the inverse dynamics of any robot. This is realized by the use of Gaussian Processes drifting parallel along the system data. An extension by a database enables the efficient use of data points from the past. The central component of this work is the implementation of such a method in a controller in order to achieve the actual goal: the feedforward control of an industrial robot by means of machine learning. This is done by splitting the procedure into two threads running parallel so that the prediction is decoupled from the computing-intensive training of the models. Experiments show that the method reduces the tracking errors more clearly than an elaborately identified rigid body model including friction. For a defined trajectory, the squared areas of the tracking errors of all axes are reduced by more than 54% compared to motion without pre-control. In addition, a highly dynamic pick-and-place experiment is used to investigate the possible changes in system dynamics. Compared to an offline trained model, the approximation error of the proposed online approach is smaller for the remaining time of the experiment after an initial phase. Furthermore, this error is smaller throughout the experiment for online learning with parallel drifting Gaussian Processes than when using a single one.

AB - The present paper deals with an online approach to learn the inverse dynamics of any robot. This is realized by the use of Gaussian Processes drifting parallel along the system data. An extension by a database enables the efficient use of data points from the past. The central component of this work is the implementation of such a method in a controller in order to achieve the actual goal: the feedforward control of an industrial robot by means of machine learning. This is done by splitting the procedure into two threads running parallel so that the prediction is decoupled from the computing-intensive training of the models. Experiments show that the method reduces the tracking errors more clearly than an elaborately identified rigid body model including friction. For a defined trajectory, the squared areas of the tracking errors of all axes are reduced by more than 54% compared to motion without pre-control. In addition, a highly dynamic pick-and-place experiment is used to investigate the possible changes in system dynamics. Compared to an offline trained model, the approximation error of the proposed online approach is smaller for the remaining time of the experiment after an initial phase. Furthermore, this error is smaller throughout the experiment for online learning with parallel drifting Gaussian Processes than when using a single one.

KW - Feedforward control

KW - Gaussian Process

KW - Implementation

KW - Inverse dynamics

KW - Online learning

UR - http://www.scopus.com/inward/record.url?scp=85072398591&partnerID=8YFLogxK

U2 - 10.1109/ICMA.2019.8816298

DO - 10.1109/ICMA.2019.8816298

M3 - Conference contribution

AN - SCOPUS:85072398591

SN - 978-1-7281-1698-3

SN - 978-1-7281-1700-3

T3 - Proceedings of 2019 IEEE International Conference on Mechatronics and Automation, ICMA 2019

SP - 962

EP - 969

BT - Proceedings of 2019 IEEE International Conference on Mechatronics and Automation, ICMA 2019

PB - Institute of Electrical and Electronics Engineers Inc.

Y2 - 4 August 2019 through 7 August 2019

ER -

Von denselben Autoren