Verification of Low-Dimensional Neural Network Control

Johan Grönqvist; Anders Rantzer

doi:10.1109/CDC49753.2023.10384149

Verification of Low-Dimensional Neural Network Control

Forskningsoutput: Kapitel i bok/rapport/Conference proceeding › Konferenspaper i proceeding › Peer review

Sammanfattning

We verify safety of a nonlinear continuous-time system controlled by a neural network controller. The system is decomposed into low-dimensional subsystems connected in a feedback loop. Our application is a rocket landing, and open-loop properties of the two-dimensional altitude subsystem are verified using worst-case simulations. Closed-loop safety properties (crash-avoidance) of the full system are obtained from composition of contracts for open-loop safety properties of subsystems in a fashion analogous to the small-gain theorem.

Originalspråk	engelska
Titel på värdpublikation	Proceedings of the IEEE Conference on Decision and Control
Förlag	IEEE - Institute of Electrical and Electronics Engineers Inc.
Sidor	4566-4571
Antal sidor	6
ISBN (elektroniskt)	9798350301243
DOI	https://doi.org/10.1109/CDC49753.2023.10384149
Status	Published - 2023
Evenemang	62nd IEEE Conference on Decision and Control, CDC 2023 - Singapore, Singapore Varaktighet: 2023 dec. 13 → 2023 dec. 15

Konferens

Konferens	62nd IEEE Conference on Decision and Control, CDC 2023
Land/Territorium	Singapore
Ort	Singapore
Period	2023/12/13 → 2023/12/15

Ämnesklassifikation (UKÄ)

Reglerteknik

Tillgång till dokument

10.1109/CDC49753.2023.10384149

1 Avslutade
1 Ej startade

Stability and Robustness with Neural Networks in Control
Grönqvist, J. (Forskare)
Projekt: Forskning
AI4GNC: Artificial intelligence techniques for guidance, navigation, and control
Cervin, A. (Forskare), Rantzer, A. (Forskare), Grönqvist, J. (Forskare), Kjellqvist, O. (Forskare), Renganathan, V. (Forskare), Vladu, E. (Forskare) & Upadhyaya, M. (Forskare)
ESA, the European Space Agency
2021/04/01 → 2022/10/31
Projekt: Uppdragsforskning

Citera det här

@inproceedings{029ebcfc55ac404dbfdcb9c63885de23,

title = "Verification of Low-Dimensional Neural Network Control",

abstract = "We verify safety of a nonlinear continuous-time system controlled by a neural network controller. The system is decomposed into low-dimensional subsystems connected in a feedback loop. Our application is a rocket landing, and open-loop properties of the two-dimensional altitude subsystem are verified using worst-case simulations. Closed-loop safety properties (crash-avoidance) of the full system are obtained from composition of contracts for open-loop safety properties of subsystems in a fashion analogous to the small-gain theorem.",

author = "Johan Gr{\"o}nqvist and Anders Rantzer",

year = "2023",

doi = "10.1109/CDC49753.2023.10384149",

language = "English",

pages = "4566--4571",

booktitle = "Proceedings of the IEEE Conference on Decision and Control",

publisher = "IEEE - Institute of Electrical and Electronics Engineers Inc.",

address = "United States",

note = "62nd IEEE Conference on Decision and Control, CDC 2023 ; Conference date: 13-12-2023 Through 15-12-2023",

}

TY - GEN

T1 - Verification of Low-Dimensional Neural Network Control

AU - Grönqvist, Johan

AU - Rantzer, Anders

PY - 2023

Y1 - 2023

N2 - We verify safety of a nonlinear continuous-time system controlled by a neural network controller. The system is decomposed into low-dimensional subsystems connected in a feedback loop. Our application is a rocket landing, and open-loop properties of the two-dimensional altitude subsystem are verified using worst-case simulations. Closed-loop safety properties (crash-avoidance) of the full system are obtained from composition of contracts for open-loop safety properties of subsystems in a fashion analogous to the small-gain theorem.

AB - We verify safety of a nonlinear continuous-time system controlled by a neural network controller. The system is decomposed into low-dimensional subsystems connected in a feedback loop. Our application is a rocket landing, and open-loop properties of the two-dimensional altitude subsystem are verified using worst-case simulations. Closed-loop safety properties (crash-avoidance) of the full system are obtained from composition of contracts for open-loop safety properties of subsystems in a fashion analogous to the small-gain theorem.

U2 - 10.1109/CDC49753.2023.10384149

DO - 10.1109/CDC49753.2023.10384149

M3 - Paper in conference proceeding

AN - SCOPUS:85184832418

SP - 4566

EP - 4571

BT - Proceedings of the IEEE Conference on Decision and Control

PB - IEEE - Institute of Electrical and Electronics Engineers Inc.

T2 - 62nd IEEE Conference on Decision and Control, CDC 2023

Y2 - 13 December 2023 through 15 December 2023

ER -

Verification of Low-Dimensional Neural Network Control

Sammanfattning

Konferens

Ämnesklassifikation (UKÄ)

Tillgång till dokument

Fingeravtryck

Projekt

Stability and Robustness with Neural Networks in Control

AI4GNC: Artificial intelligence techniques for guidance, navigation, and control

Citera det här