---
_id: '59740'
abstract:
- lang: eng
text: ABSTRACTIn this contribution, we propose
an innovative method for determining optimal control sequences for nonlinear systems
with partially unknown dynamics, which further expands our previous work. Within
the paradigm of model‐based design, the practicality and safety of commissioning
feedforward controls and feedback controllers have priority. Our approach leverages
probabilistic Gaussian processes to adjust for model inaccuracies from measured
system data. This differs from conventional approaches that involve complicated
analytical modeling and may entail a substantial time investment to acquire expertise
and may prove impractical. Consequently, we address the limitations inherent in
traditional design methodologies. Our research focuses on the formulation and
solution of the hybrid1 optimal control problem using probabilistic
state predictions and multiple shooting. This ensures adaptability, data efficiency,
and resilience against uncertainties in system dynamics. These attributes are
empirically substantiated through experimental validation on a chaotic and highly
sensitive dynamical system—a double pendulum on a cart. Our methodology unfolds
as an iterative learning process, systematically exploring diverse controls, accumulating
data within each iteration, and refining the control strategy until the desired
task is accomplished. The adoption of the two‐degree‐of‐freedom control structure
allows for the distinct consideration of the feedforward and the feedback control
signal. For the latter, we employ a time‐variant, linear quadratic regulator (LQR)
designed to stabilize the system around its target trajectory. Furthermore, we
integrate a probabilistic long‐term prediction through the unscented transform,
enabling systematic anticipation of safety‐critical violations. Detailed insights
into relevant implementation aspects are provided. To ascertain the real‐world
applicability, we present an exemplary application involving a double pendulum
on a cart. The objective is to bring the pendulum arms from the lower stable to
the upper unstable equilibrium by horizontally moving the cart and subsequently
stabilize them. In this scenario, we assume that the centrifugal forces, crucial
to the system dynamics, have not been accurately modeled and must be learned from
data. Solving the control task took only 5 iterations and 1 h of computation time,
which surpasses our previous work [2], where we used the purely data‐driven PILCO
framework and required 27 iterations and 57 h of computation time. The time of
interaction with the system decreased by and the computation time is lowered
by . It demonstrates significant practical applicability for commissioning control systems.
author:
- first_name: Michael
full_name: Hesse, Michael
id: '29222'
last_name: Hesse
- first_name: Luis
full_name: Schwarzer, Luis
last_name: Schwarzer
- first_name: Julia
full_name: Timmermann, Julia
id: '15402'
last_name: Timmermann
- first_name: Ansgar
full_name: Trächtler, Ansgar
id: '552'
last_name: Trächtler
citation:
ama: Hesse M, Schwarzer L, Timmermann J, Trächtler A. Robust and Efficient Hybrid
Optimal Control via Gaussian Process Regression and Multiple Shooting With Experimental
Validation on a Double Pendulum on a Cart. PAMM. 2025;25(2). doi:10.1002/pamm.70004
apa: Hesse, M., Schwarzer, L., Timmermann, J., & Trächtler, A. (2025). Robust
and Efficient Hybrid Optimal Control via Gaussian Process Regression and Multiple
Shooting With Experimental Validation on a Double Pendulum on a Cart. PAMM,
25(2). https://doi.org/10.1002/pamm.70004
bibtex: '@article{Hesse_Schwarzer_Timmermann_Trächtler_2025, title={Robust and Efficient
Hybrid Optimal Control via Gaussian Process Regression and Multiple Shooting With
Experimental Validation on a Double Pendulum on a Cart}, volume={25}, DOI={10.1002/pamm.70004}, number={2},
journal={PAMM}, publisher={Wiley}, author={Hesse, Michael and Schwarzer, Luis
and Timmermann, Julia and Trächtler, Ansgar}, year={2025} }'
chicago: Hesse, Michael, Luis Schwarzer, Julia Timmermann, and Ansgar Trächtler.
“Robust and Efficient Hybrid Optimal Control via Gaussian Process Regression and
Multiple Shooting With Experimental Validation on a Double Pendulum on a Cart.”
PAMM 25, no. 2 (2025). https://doi.org/10.1002/pamm.70004.
ieee: 'M. Hesse, L. Schwarzer, J. Timmermann, and A. Trächtler, “Robust and Efficient
Hybrid Optimal Control via Gaussian Process Regression and Multiple Shooting With
Experimental Validation on a Double Pendulum on a Cart,” PAMM, vol. 25,
no. 2, 2025, doi: 10.1002/pamm.70004.'
mla: Hesse, Michael, et al. “Robust and Efficient Hybrid Optimal Control via Gaussian
Process Regression and Multiple Shooting With Experimental Validation on a Double
Pendulum on a Cart.” PAMM, vol. 25, no. 2, Wiley, 2025, doi:10.1002/pamm.70004.
short: M. Hesse, L. Schwarzer, J. Timmermann, A. Trächtler, PAMM 25 (2025).
date_created: 2025-04-30T08:18:46Z
date_updated: 2025-09-03T10:35:24Z
department:
- _id: '880'
- _id: '153'
doi: 10.1002/pamm.70004
intvolume: ' 25'
issue: '2'
language:
- iso: eng
project:
- _id: '690'
name: 'DART: Datengetriebene Methoden in der Regelungstechnik'
publication: PAMM
publication_identifier:
issn:
- 1617-7061
- 1617-7061
publication_status: published
publisher: Wiley
status: public
title: Robust and Efficient Hybrid Optimal Control via Gaussian Process Regression
and Multiple Shooting With Experimental Validation on a Double Pendulum on a Cart
type: journal_article
user_id: '15402'
volume: 25
year: '2025'
...
---
_id: '61118'
abstract:
- lang: eng
text: Im Zuge der Digitalisierung erfahren maschinelles Lernen und datengetriebene
Methoden derzeit eine große Aufmerksamkeit in Wissenschaft und Industrie. Es fehlt
jedoch an Grundlagenwissen und Verständnis, wie die datengetriebenen Methoden
der Informatik mit bewährten modellbasierten Ingenieursmethoden wie dem modellbasierten
Entwurf in der Mechatronik und Methoden der Regelungstechnik sinnvoll kombiniert
werden können, um hybride Modelle zu erhalten. Diese ingenieurwissenschaftlichen
Methoden basieren auf physikalischen Verhaltensmodellen, die eine besonders verdichtete
und interpretierbare Darstellung von Wissen darstellen und insbesondere kausale
Zusammenhänge beschreiben. Für spezifische regelungstechnische Anwendungen gibt
es umfangreiches Vorwissen in Form von bekannten Strukturen und Informationen,
wie z.B. (Teil-)Modelle oder Parametersätze, die auch bei der Anwendung von Methoden
wie dem maschinellen Lernen genutzt werden sollten. Eine solche sinnvolle systematische
Verknüpfung ist wissenschaftlich, insbesondere im Hinblick auf die industrielle
Anwendung, noch nicht ausreichend untersucht worden und sehr vielversprechend.
In diesem Beitrag werden die Ergebnisse der Nachwuchsforschungsgruppe DART – Datengetriebene
Methoden in der Regelungstechnik vorgestellt. Das Hauptziel war es, die synergetische
Kombination von modell- und datengetriebenen Methoden für regelungstechnische
Aufgaben zu erforschen und es werden alle wichtigen Forschungsergebnisse aber
auch die verwendeten Grundprinzipien des maschinellen Lernens in diesem Beitrag
dargestellt.
- lang: eng
text: In the course of digitalization, machine learning and data-driven methods
are currently receiving a great deal of attention in science and industry. However,
there is a lack of basic knowledge and understanding of how data-driven methods
in computer science can be meaningfully combined with proven model-based engineering
methods such as model-based design in mechatronics and control engineering methods
to obtain hybrid models. These engineering methods are based on physical models,
which represent a particularly condensed and interpretable representation of knowledge
and, in particular, describe causal relationships. For specific control engineering
applications, there is extensive prior knowledge in the form of known structures
and information, such as (partial) models or parameter sets, which should also
be used when applying methods such as machine learning. Such a meaningful systematic
connection has not yet been sufficiently investigated scientifically, especially
with regard to industrial applications, and is very promising. This contribution
presents the results of the DART junior research group – Data-driven methods in
control engineering. The main objective was to investigate the synergistic combination
of model- and data-driven methods for control engineering tasks, and all important
research results as well as the basic principles of machine learning used are
presented in this publication.
author:
- first_name: Julia
full_name: Timmermann, Julia
id: '15402'
last_name: Timmermann
- first_name: Ricarda-Samantha
full_name: Götte, Ricarda-Samantha
id: '43992'
last_name: Götte
- first_name: Annika
full_name: Junker, Annika
id: '41470'
last_name: Junker
orcid: 0009-0002-6475-2503
- first_name: Michael
full_name: Hesse, Michael
id: '29222'
last_name: Hesse
- first_name: Luis
full_name: Schwarzer, Luis
last_name: Schwarzer
citation:
ama: Timmermann J, Götte R-S, Junker A, Hesse M, Schwarzer L. DART - Datengetriebene
Methoden in der Regelungstechnik. Vol Band 430. 1. Auflage. HNI Verlagsschriftenreihe;
2025. doi:10.17619/UNIPB/1-2305
apa: 'Timmermann, J., Götte, R.-S., Junker, A., Hesse, M., & Schwarzer, L. (2025).
DART - Datengetriebene Methoden in der Regelungstechnik: Vol. Band 430
(1. Auflage). HNI Verlagsschriftenreihe. https://doi.org/10.17619/UNIPB/1-2305'
bibtex: '@book{Timmermann_Götte_Junker_Hesse_Schwarzer_2025, place={Paderborn},
edition={1. Auflage}, title={DART - Datengetriebene Methoden in der Regelungstechnik},
volume={Band 430}, DOI={10.17619/UNIPB/1-2305},
publisher={HNI Verlagsschriftenreihe}, author={Timmermann, Julia and Götte, Ricarda-Samantha
and Junker, Annika and Hesse, Michael and Schwarzer, Luis}, year={2025} }'
chicago: 'Timmermann, Julia, Ricarda-Samantha Götte, Annika Junker, Michael Hesse,
and Luis Schwarzer. DART - Datengetriebene Methoden in der Regelungstechnik.
1. Auflage. Vol. Band 430. Paderborn: HNI Verlagsschriftenreihe, 2025. https://doi.org/10.17619/UNIPB/1-2305.'
ieee: 'J. Timmermann, R.-S. Götte, A. Junker, M. Hesse, and L. Schwarzer, DART
- Datengetriebene Methoden in der Regelungstechnik, 1. Auflage., vol. Band
430. Paderborn: HNI Verlagsschriftenreihe, 2025.'
mla: Timmermann, Julia, et al. DART - Datengetriebene Methoden in der Regelungstechnik.
1. Auflage, vol. Band 430, HNI Verlagsschriftenreihe, 2025, doi:10.17619/UNIPB/1-2305.
short: J. Timmermann, R.-S. Götte, A. Junker, M. Hesse, L. Schwarzer, DART - Datengetriebene
Methoden in der Regelungstechnik, 1. Auflage, HNI Verlagsschriftenreihe, Paderborn,
2025.
date_created: 2025-09-03T09:35:35Z
date_updated: 2026-04-01T06:14:00Z
department:
- _id: '880'
- _id: '153'
doi: 10.17619/UNIPB/1-2305
edition: 1. Auflage
language:
- iso: ger
main_file_link:
- open_access: '1'
url: https://digital.ub.uni-paderborn.de/doi/10.17619/UNIPB/1-2305
oa: '1'
place: Paderborn
publication_status: published
publisher: HNI Verlagsschriftenreihe
status: public
title: DART - Datengetriebene Methoden in der Regelungstechnik
type: book
user_id: '41470'
volume: Band 430
year: '2025'
...
---
_id: '58448'
abstract:
- lang: eng
text: Die Inbetriebnahme von Steuerungen und Regelungen stellt sicher, dass ein
mechatronisches System ordnungsgemäß funktioniert und den Anforderungen gerecht
wird. Der modellbasierte Entwurf basiert auf einem genauen Simulationsmodell.
Allerdings ist dieser klassische Weg bei komplexen Systemen oft nicht praktikabel,
da die analytische Modellierung zu kompliziert und zeitaufwendig ist. Diese Forschungslücke
wird durch Verfahren adressiert, die eine effiziente und sichere Inbetriebnahme
ermöglichen. Diese Verfahren kombinieren Regelungstechnik und Reinforcement Learning
und nutzen vorhandenes Wissen über die Regelungsaufgabe, um Korrekturen basierend
auf Messdaten und der probabilistischen Gauß-Prozess-Regression vorzunehmen. Das
Vorwissen kann als teilweise bekanntes physikalisches Modell oder als Steuerungsfunktion
vorliegen. Anwendungsbeispiele sind der Ultraschalldrahtbondprozess, verschiedene
Pendelsysteme und ein Hexapod. Eine angepasste Bayessche Optimierung wird zur
Identifikation einer Steuerparametrisierung für das Ultraschallbonden eingesetzt.
Außerdem wird eine hybride Optimalsteuerung für das Doppelpendel auf einem Wagen
entwickelt und erfolgreich validiert. Fur einen Hexapod zur Fahrzeugachsprüfung
wird eine hybride Zustandslinearisierung formuliert und ein Funktionsnachweis
im Rahmen einer Simulation erbracht. Die Einhaltung technischer Rahmenbedingungen
und stabiles Systemverhalten werden durch probabilistische Pradiktionen gewährleistet.
In allen Anwendungsfällen wird eine Steigerung der Effizienz und Güte erzielt.
- lang: eng
text: The commissioning of control systems ensures that a mechatronic system functions
properly and meets the requirements. Model-based design is based on a precise
simulation model. However, this classic approach is often impractical for complex
systems, as analytical modeling is too complicated and time-consuming. This research
gap is addressed by methods that enable efficient and safe commissioning. These
methods combine control engineering and reinforcement learning and use existing
knowledge about the control task to make corrections based on measurement data
and probabilistic Gaussian process regression. The prior knowledge can be available
as a partially known physical model or as a control function. Application examples
include the ultrasonic wire bonding process, various pendulum systems and a hexapod.
An adapted Bayesian optimization is used to identify a control parameterization
for ultrasonic bonding. In addition, a hybrid optimal control for the double pendulum
on a cart is developed and successfully validated. A hybrid state linearization
is formulated for a hexapod for vehicle axle testing and a proof of concept is
provided in a simulation. Compliance with technical framework conditions and stable
system behavior are ensured by probabilistic predictions. An increase in efficiency
and quality is achieved in all use cases.
author:
- first_name: Michael
full_name: Hesse, Michael
id: '29222'
last_name: Hesse
citation:
ama: Hesse M. Interaktive Inbetriebnahme von Steuerungen und Regelungen für partiell
bekannte dynamische Systeme mittels Gauß-Prozess-Regression. Vol 426. Heinz
Nixdorf Institut; 2024. doi:10.17619/UNIPB/1-2135
apa: Hesse, M. (2024). Interaktive Inbetriebnahme von Steuerungen und Regelungen
für partiell bekannte dynamische Systeme mittels Gauß-Prozess-Regression (Vol.
426). Heinz Nixdorf Institut. https://doi.org/10.17619/UNIPB/1-2135
bibtex: '@book{Hesse_2024, place={Paderborn}, series={Verlagsschriftenreihe des
Heinz Nixdorf Instituts}, title={Interaktive Inbetriebnahme von Steuerungen und
Regelungen für partiell bekannte dynamische Systeme mittels Gauß-Prozess-Regression},
volume={426}, DOI={10.17619/UNIPB/1-2135},
publisher={Heinz Nixdorf Institut}, author={Hesse, Michael}, year={2024}, collection={Verlagsschriftenreihe
des Heinz Nixdorf Instituts} }'
chicago: 'Hesse, Michael. Interaktive Inbetriebnahme von Steuerungen und Regelungen
für partiell bekannte dynamische Systeme mittels Gauß-Prozess-Regression.
Vol. 426. Verlagsschriftenreihe des Heinz Nixdorf Instituts. Paderborn: Heinz
Nixdorf Institut, 2024. https://doi.org/10.17619/UNIPB/1-2135.'
ieee: 'M. Hesse, Interaktive Inbetriebnahme von Steuerungen und Regelungen für
partiell bekannte dynamische Systeme mittels Gauß-Prozess-Regression, vol.
426. Paderborn: Heinz Nixdorf Institut, 2024.'
mla: Hesse, Michael. Interaktive Inbetriebnahme von Steuerungen und Regelungen
für partiell bekannte dynamische Systeme mittels Gauß-Prozess-Regression.
Heinz Nixdorf Institut, 2024, doi:10.17619/UNIPB/1-2135.
short: M. Hesse, Interaktive Inbetriebnahme von Steuerungen und Regelungen für partiell
bekannte dynamische Systeme mittels Gauß-Prozess-Regression, Heinz Nixdorf Institut,
Paderborn, 2024.
date_created: 2025-01-30T14:45:46Z
date_updated: 2025-01-30T14:58:46Z
department:
- _id: '153'
- _id: '880'
doi: 10.17619/UNIPB/1-2135
intvolume: ' 426'
language:
- iso: ger
main_file_link:
- open_access: '1'
url: https://digital.ub.uni-paderborn.de/doi/10.17619/UNIPB/1-2135
oa: '1'
place: Paderborn
publication_identifier:
eissn:
- 2365-4422
isbn:
- 978-3-947647-45-3
publication_status: published
publisher: Heinz Nixdorf Institut
series_title: Verlagsschriftenreihe des Heinz Nixdorf Instituts
status: public
supervisor:
- first_name: Julia
full_name: Timmermann, Julia
id: '15402'
last_name: Timmermann
- first_name: Eyke
full_name: Hüllermeier, Eyke
last_name: Hüllermeier
title: Interaktive Inbetriebnahme von Steuerungen und Regelungen für partiell bekannte
dynamische Systeme mittels Gauß-Prozess-Regression
type: dissertation
user_id: '82875'
volume: 426
year: '2024'
...
---
_id: '48476'
author:
- first_name: Michael
full_name: Hesse, Michael
id: '29222'
last_name: Hesse
- first_name: Julia
full_name: Timmermann, Julia
id: '15402'
last_name: Timmermann
- first_name: Ansgar
full_name: Trächtler, Ansgar
id: '552'
last_name: Trächtler
citation:
ama: 'Hesse M, Timmermann J, Trächtler A. Hybrid Optimal Control for Dynamical Systems
using Gaussian Process Regression and Unscented Transform*. In: 2023
European Control Conference (ECC). IEEE; 2023. doi:10.23919/ecc57647.2023.10178368'
apa: Hesse, M., Timmermann, J., & Trächtler, A. (2023). Hybrid Optimal Control
for Dynamical Systems using Gaussian Process Regression and Unscented Transform*.
2023 European Control Conference (ECC). https://doi.org/10.23919/ecc57647.2023.10178368
bibtex: '@inproceedings{Hesse_Timmermann_Trächtler_2023, title={Hybrid Optimal Control
for Dynamical Systems using Gaussian Process Regression and Unscented Transform*},
DOI={10.23919/ecc57647.2023.10178368},
booktitle={2023 European Control Conference (ECC)}, publisher={IEEE}, author={Hesse,
Michael and Timmermann, Julia and Trächtler, Ansgar}, year={2023} }'
chicago: Hesse, Michael, Julia Timmermann, and Ansgar Trächtler. “Hybrid Optimal
Control for Dynamical Systems Using Gaussian Process Regression and Unscented
Transform*.” In 2023 European Control Conference (ECC). IEEE,
2023. https://doi.org/10.23919/ecc57647.2023.10178368.
ieee: 'M. Hesse, J. Timmermann, and A. Trächtler, “Hybrid Optimal Control for Dynamical
Systems using Gaussian Process Regression and Unscented Transform*,”
2023, doi: 10.23919/ecc57647.2023.10178368.'
mla: Hesse, Michael, et al. “Hybrid Optimal Control for Dynamical Systems Using
Gaussian Process Regression and Unscented Transform*.” 2023 European
Control Conference (ECC), IEEE, 2023, doi:10.23919/ecc57647.2023.10178368.
short: 'M. Hesse, J. Timmermann, A. Trächtler, in: 2023 European Control Conference
(ECC), IEEE, 2023.'
date_created: 2023-10-25T13:56:34Z
date_updated: 2024-11-13T08:43:40Z
department:
- _id: '153'
- _id: '880'
doi: 10.23919/ecc57647.2023.10178368
language:
- iso: eng
publication: 2023 European Control Conference (ECC)
publication_status: published
publisher: IEEE
quality_controlled: '1'
status: public
title: Hybrid Optimal Control for Dynamical Systems using Gaussian Process Regression
and Unscented Transform*
type: conference
user_id: '82875'
year: '2023'
...
---
_id: '29803'
abstract:
- lang: eng
text: "Ultrasonic wire bonding is a solid-state joining process used to form electrical
interconnections in micro and\r\npower electronics and batteries. A high frequency
oscillation causes a metallurgical bond deformation in\r\nthe contact area. Due
to the numerous physical influencing factors, it is very difficult to accurately
capture\r\nthis process in a model. Therefore, our goal is to determine a suitable
feed-forward control strategy for the\r\nbonding process even without detailed
model knowledge. We propose the use of batch constrained Bayesian\r\noptimization
for the control design. Hence, Bayesian optimization is precisely adapted to the
application of\r\nbonding: the constraint is used to check one quality feature
of the process and the use of batches leads to\r\nmore efficient experiments.
Our approach is suitable to determine a feed-forward control for the bonding\r\nprocess
that provides very high quality bonds without using a physical model. We also
show that the quality\r\nof the Bayesian optimization based control outperforms
random search as well as manual search by a user.\r\nUsing a simple prior knowledge
model derived from data further improves the quality of the connection.\r\nThe
Bayesian optimization approach offers the possibility to perform a sensitivity
analysis of the control\r\nparameters, which allows to evaluate the influence
of each control parameter on the bond quality. In summary,\r\nBayesian optimization
applied to the bonding process provides an excellent opportunity to develop a
feedforward\r\ncontrol without full modeling of the underlying physical processes."
author:
- first_name: Michael
full_name: Hesse, Michael
id: '29222'
last_name: Hesse
- first_name: Matthias
full_name: Hunstig, Matthias
last_name: Hunstig
- first_name: Julia
full_name: Timmermann, Julia
id: '15402'
last_name: Timmermann
- first_name: Ansgar
full_name: Trächtler, Ansgar
id: '552'
last_name: Trächtler
citation:
ama: 'Hesse M, Hunstig M, Timmermann J, Trächtler A. Batch Constrained Bayesian
Optimization for UltrasonicWire Bonding Feed-forward Control Design. In: Proceedings
of the 11th International Conference on Pattern Recognition Applications and Methods
(ICPRAM). ; 2022:383-394.'
apa: Hesse, M., Hunstig, M., Timmermann, J., & Trächtler, A. (2022). Batch Constrained
Bayesian Optimization for UltrasonicWire Bonding Feed-forward Control Design.
Proceedings of the 11th International Conference on Pattern Recognition Applications
and Methods (ICPRAM), 383–394.
bibtex: '@inproceedings{Hesse_Hunstig_Timmermann_Trächtler_2022, title={Batch Constrained
Bayesian Optimization for UltrasonicWire Bonding Feed-forward Control Design},
booktitle={Proceedings of the 11th International Conference on Pattern Recognition
Applications and Methods (ICPRAM)}, author={Hesse, Michael and Hunstig, Matthias
and Timmermann, Julia and Trächtler, Ansgar}, year={2022}, pages={383–394} }'
chicago: Hesse, Michael, Matthias Hunstig, Julia Timmermann, and Ansgar Trächtler.
“Batch Constrained Bayesian Optimization for UltrasonicWire Bonding Feed-Forward
Control Design.” In Proceedings of the 11th International Conference on Pattern
Recognition Applications and Methods (ICPRAM), 383–94, 2022.
ieee: M. Hesse, M. Hunstig, J. Timmermann, and A. Trächtler, “Batch Constrained
Bayesian Optimization for UltrasonicWire Bonding Feed-forward Control Design,”
in Proceedings of the 11th International Conference on Pattern Recognition
Applications and Methods (ICPRAM), Online, 2022, pp. 383–394.
mla: Hesse, Michael, et al. “Batch Constrained Bayesian Optimization for UltrasonicWire
Bonding Feed-Forward Control Design.” Proceedings of the 11th International
Conference on Pattern Recognition Applications and Methods (ICPRAM), 2022,
pp. 383–94.
short: 'M. Hesse, M. Hunstig, J. Timmermann, A. Trächtler, in: Proceedings of the
11th International Conference on Pattern Recognition Applications and Methods
(ICPRAM), 2022, pp. 383–394.'
conference:
end_date: 2022-02-05
location: Online
name: 11th International Conference on Pattern Recognition Applications and Methods
start_date: 2022-02-03
date_created: 2022-02-09T12:50:25Z
date_updated: 2024-11-13T08:44:17Z
department:
- _id: '153'
- _id: '880'
keyword:
- Bayesian optimization
- Wire bonding
- Feed-forward control
- model-free design
language:
- iso: eng
page: 383-394
publication: Proceedings of the 11th International Conference on Pattern Recognition
Applications and Methods (ICPRAM)
publication_identifier:
isbn:
- 978-989-758-549-4
quality_controlled: '1'
status: public
title: Batch Constrained Bayesian Optimization for UltrasonicWire Bonding Feed-forward
Control Design
type: conference
user_id: '82875'
year: '2022'
...
---
_id: '22996'
abstract:
- lang: eng
text: The effective control design of a dynamical system traditionally relies on
a high level of system understanding, usually expressed in terms of an exact physical
model. In contrast to this, reinforcement learning adopts a data-driven approach
and constructs an optimal control strategy by interacting with the underlying
system. To keep the wear of real-world systems as low as possible, the learning
process should be short. In our research, we used the state-of-the-art reinforcement
learning method PILCO to design a feedback control strategy for the swing-up of
the double pendulum on a cart with remarkably few test iterations at the test
bench. PILCO stands for “probabilistic inference for learning control” and requires
only few expert knowledge for learning. To achieve the swing-up of a double pendulum
on a cart to its upper unstable equilibrium position, we introduce additional
state restrictions to PILCO, so that the limited cart distance can be taken into
account. Thanks to these measures, we were able to learn the swing up at the real
test bench for the first time and in only 27 learning iterations.
author:
- first_name: Michael
full_name: Hesse, Michael
id: '29222'
last_name: Hesse
- first_name: Julia
full_name: Timmermann, Julia
id: '15402'
last_name: Timmermann
- first_name: Eyke
full_name: Hüllermeier, Eyke
id: '48129'
last_name: Hüllermeier
- first_name: Ansgar
full_name: Trächtler, Ansgar
id: '552'
last_name: Trächtler
citation:
ama: Hesse M, Timmermann J, Hüllermeier E, Trächtler A. A Reinforcement Learning
Strategy for the Swing-Up of the Double Pendulum on a Cart. Procedia Manufacturing.
2018;24:15-20.
apa: Hesse, M., Timmermann, J., Hüllermeier, E., & Trächtler, A. (2018). A Reinforcement
Learning Strategy for the Swing-Up of the Double Pendulum on a Cart. Procedia
Manufacturing, 24, 15–20.
bibtex: '@article{Hesse_Timmermann_Hüllermeier_Trächtler_2018, title={A Reinforcement
Learning Strategy for the Swing-Up of the Double Pendulum on a Cart}, volume={24},
journal={Procedia Manufacturing}, author={Hesse, Michael and Timmermann, Julia
and Hüllermeier, Eyke and Trächtler, Ansgar}, year={2018}, pages={15–20} }'
chicago: 'Hesse, Michael, Julia Timmermann, Eyke Hüllermeier, and Ansgar Trächtler.
“A Reinforcement Learning Strategy for the Swing-Up of the Double Pendulum on
a Cart.” Procedia Manufacturing 24 (2018): 15–20.'
ieee: M. Hesse, J. Timmermann, E. Hüllermeier, and A. Trächtler, “A Reinforcement
Learning Strategy for the Swing-Up of the Double Pendulum on a Cart,” Procedia
Manufacturing, vol. 24, pp. 15–20, 2018.
mla: Hesse, Michael, et al. “A Reinforcement Learning Strategy for the Swing-Up
of the Double Pendulum on a Cart.” Procedia Manufacturing, vol. 24, 2018,
pp. 15–20.
short: M. Hesse, J. Timmermann, E. Hüllermeier, A. Trächtler, Procedia Manufacturing
24 (2018) 15–20.
date_created: 2021-08-09T05:41:38Z
date_updated: 2023-11-06T15:17:24Z
department:
- _id: '153'
intvolume: ' 24'
language:
- iso: eng
page: 15 - 20
publication: Procedia Manufacturing
quality_controlled: '1'
status: public
title: A Reinforcement Learning Strategy for the Swing-Up of the Double Pendulum on
a Cart
type: journal_article
user_id: '29222'
volume: 24
year: '2018'
...