Physics-Guided Neural Networks for Feedforward Control: An Orthogonal Projection-Based Approach

Johan Kon; Dennis Bruijnen; Jeroen  van de Wijdeven; Marcel  Heertjes; T.A.E. Oomen

doi:10.23919/ACC53348.2022.9867653

Physics-Guided Neural Networks for Feedforward Control: An Orthogonal Projection-Based Approach

Johan Kon, Dennis Bruijnen, Jeroen van de Wijdeven, Marcel Heertjes, T.A.E. Oomen

Team Jan-Willem van Wingerden

Research output: Chapter in Book/Conference proceedings/Edited volume › Conference contribution › Scientific › peer-review

11 Citations (Scopus)

12 Downloads (Pure)

Abstract

Unknown nonlinear dynamics can limit the performance of model-based feedforward control. The aim of this paper is to develop a feedforward control framework for systems with unknown, typically nonlinear, dynamics. To address the unknown dynamics, a physics-based feedforward model is complemented by a neural network. The neural network output in the subspace of the model is penalized through orthogonal projection. This results in uniquely identifiable model coefficients, enabling increased performance and similar task flexibility with respect to the model-based controller. The feedforward framework is validated on a representative system with performance limiting nonlinear friction characteristics.

Original language	English
Title of host publication	Proceedings of the American Control Conference (ACC 2022)
Publisher	IEEE
Pages	4377-4382
ISBN (Print)	978-1-6654-5196-3
DOIs	https://doi.org/10.23919/ACC53348.2022.9867653
Publication status	Published - 2022
Event	2022 American Control Conference, ACC 2022 - Atlanta, United States Duration: 8 Jun 2022 → 10 Jun 2022

Conference

Conference	2022 American Control Conference, ACC 2022
Country/Territory	United States
City	Atlanta
Period	8/06/22 → 10/06/22

Bibliographical note

Green Open Access added to TU Delft Institutional Repository 'You share, we take care!' - Taverne project https://www.openaccess.nl/en/you-share-we-take-care
Otherwise as indicated in the copyright section: the publisher is the copyright holder of this work and the author uses the Dutch legislation to make this work public.

Keywords

Training
Maximum likelihood detection
Limiting
System dynamics
Nonlinear filters
Cost function
Feedforward neural networks

Access to Document

10.23919/ACC53348.2022.9867653

Physics-Guided_Neural_Networks_for_Feedforward_Control_An_Orthogonal_Projection-Based_ApproachFinal published version, 400 KB

Cite this

@inproceedings{b0513881b2e74985bd220f49a121e6b8,

title = "Physics-Guided Neural Networks for Feedforward Control: An Orthogonal Projection-Based Approach",

abstract = "Unknown nonlinear dynamics can limit the performance of model-based feedforward control. The aim of this paper is to develop a feedforward control framework for systems with unknown, typically nonlinear, dynamics. To address the unknown dynamics, a physics-based feedforward model is complemented by a neural network. The neural network output in the subspace of the model is penalized through orthogonal projection. This results in uniquely identifiable model coefficients, enabling increased performance and similar task flexibility with respect to the model-based controller. The feedforward framework is validated on a representative system with performance limiting nonlinear friction characteristics.",

keywords = "Training, Maximum likelihood detection, Limiting, System dynamics, Nonlinear filters, Cost function, Feedforward neural networks",

author = "Johan Kon and Dennis Bruijnen and {van de Wijdeven}, Jeroen and Marcel Heertjes and T.A.E. Oomen",

note = "Green Open Access added to TU Delft Institutional Repository 'You share, we take care!' - Taverne project https://www.openaccess.nl/en/you-share-we-take-care Otherwise as indicated in the copyright section: the publisher is the copyright holder of this work and the author uses the Dutch legislation to make this work public.; 2022 American Control Conference, ACC 2022 ; Conference date: 08-06-2022 Through 10-06-2022",

year = "2022",

doi = "10.23919/ACC53348.2022.9867653",

language = "English",

isbn = "978-1-6654-5196-3",

pages = "4377--4382",

booktitle = "Proceedings of the American Control Conference (ACC 2022)",

publisher = "IEEE",

address = "United States",

}

Kon, J, Bruijnen, D, van de Wijdeven, J, Heertjes, M & Oomen, TAE 2022, Physics-Guided Neural Networks for Feedforward Control: An Orthogonal Projection-Based Approach. in Proceedings of the American Control Conference (ACC 2022). IEEE, pp. 4377-4382, 2022 American Control Conference, ACC 2022, Atlanta, United States, 8/06/22. https://doi.org/10.23919/ACC53348.2022.9867653

Physics-Guided Neural Networks for Feedforward Control: An Orthogonal Projection-Based Approach. / Kon, Johan; Bruijnen, Dennis; van de Wijdeven, Jeroen et al.
Proceedings of the American Control Conference (ACC 2022). IEEE, 2022. p. 4377-4382.

Research output: Chapter in Book/Conference proceedings/Edited volume › Conference contribution › Scientific › peer-review

TY - GEN

T1 - Physics-Guided Neural Networks for Feedforward Control

T2 - 2022 American Control Conference, ACC 2022

AU - Kon, Johan

AU - Bruijnen, Dennis

AU - van de Wijdeven, Jeroen

AU - Heertjes, Marcel

AU - Oomen, T.A.E.

N1 - Green Open Access added to TU Delft Institutional Repository 'You share, we take care!' - Taverne project https://www.openaccess.nl/en/you-share-we-take-care Otherwise as indicated in the copyright section: the publisher is the copyright holder of this work and the author uses the Dutch legislation to make this work public.

PY - 2022

Y1 - 2022

N2 - Unknown nonlinear dynamics can limit the performance of model-based feedforward control. The aim of this paper is to develop a feedforward control framework for systems with unknown, typically nonlinear, dynamics. To address the unknown dynamics, a physics-based feedforward model is complemented by a neural network. The neural network output in the subspace of the model is penalized through orthogonal projection. This results in uniquely identifiable model coefficients, enabling increased performance and similar task flexibility with respect to the model-based controller. The feedforward framework is validated on a representative system with performance limiting nonlinear friction characteristics.

AB - Unknown nonlinear dynamics can limit the performance of model-based feedforward control. The aim of this paper is to develop a feedforward control framework for systems with unknown, typically nonlinear, dynamics. To address the unknown dynamics, a physics-based feedforward model is complemented by a neural network. The neural network output in the subspace of the model is penalized through orthogonal projection. This results in uniquely identifiable model coefficients, enabling increased performance and similar task flexibility with respect to the model-based controller. The feedforward framework is validated on a representative system with performance limiting nonlinear friction characteristics.

KW - Training

KW - Maximum likelihood detection

KW - Limiting

KW - System dynamics

KW - Nonlinear filters

KW - Cost function

KW - Feedforward neural networks

UR - http://www.scopus.com/inward/record.url?scp=85133681464&partnerID=8YFLogxK

U2 - 10.23919/ACC53348.2022.9867653

DO - 10.23919/ACC53348.2022.9867653

M3 - Conference contribution

SN - 978-1-6654-5196-3

SP - 4377

EP - 4382

BT - Proceedings of the American Control Conference (ACC 2022)

PB - IEEE

Y2 - 8 June 2022 through 10 June 2022

ER -

Physics-Guided Neural Networks for Feedforward Control: An Orthogonal Projection-Based Approach

Abstract

Conference

Bibliographical note

Keywords

Access to Document

Other files and links

Fingerprint

Cite this