Skip to main content

Lund University Publications

LUND UNIVERSITY LIBRARIES

Accelerated Deep Reinforcement Learning for Fast Feedback of Beam Dynamics at KARA

Wang, Weija ; Caselle, Michele ; Boltz, Tobias ; Blomley, Edmund ; Brosi, Miriam LU orcid ; Dritschler, Timo ; Ebersoldt, Andreas ; Kopmann, Andreas ; Garcia, Andrea Santamaria and Schreiber, Patrick , et al. (2021) In IEEE Transactions on Nuclear Science 68(8). p.1794-1800
Abstract

Coherent synchrotron radiation (CSR) is generated when the electron bunch length is in the order of the magnitude of the wavelength of the emitted radiation. The self-interaction of short electron bunches with their own electromagnetic fields changes the longitudinal beam dynamics significantly. Above a certain current threshold, the micro-bunching instability develops, characterized by the appearance of distinguishable substructures in the longitudinal phase space of the bunch. To stabilize the CSR emission, a real-time feedback control loop based on reinforcement learning (RL) is proposed. Informed by the available THz diagnostics, the feedback is designed to act on the radio frequency (RF) system of the storage ring to mitigate the... (More)

Coherent synchrotron radiation (CSR) is generated when the electron bunch length is in the order of the magnitude of the wavelength of the emitted radiation. The self-interaction of short electron bunches with their own electromagnetic fields changes the longitudinal beam dynamics significantly. Above a certain current threshold, the micro-bunching instability develops, characterized by the appearance of distinguishable substructures in the longitudinal phase space of the bunch. To stabilize the CSR emission, a real-time feedback control loop based on reinforcement learning (RL) is proposed. Informed by the available THz diagnostics, the feedback is designed to act on the radio frequency (RF) system of the storage ring to mitigate the micro-bunching dynamics. To satisfy low-latency requirements given by the longitudinal beam dynamics, the RL controller has been implemented on hardware (FPGA). In this article, a real-time feedback loop architecture and its performance is presented and compared with a software implementation using Keras-RL on CPU/GPU. The results obtained with the CSR simulation Inovesa demonstrate that the functionality of both platforms is equivalent. The training performance of the hardware implementation is similar to software solution, while it outperforms the Keras-RL implementation by an order of magnitude. The presented RL hardware controller is considered as an essential platform for the development of intelligent CSR control systems.

(Less)
Please use this url to cite or link to this publication:
author
; ; ; ; ; ; ; ; and , et al. (More)
; ; ; ; ; ; ; ; ; ; ; ; and (Less)
publishing date
type
Contribution to journal
publication status
published
keywords
Artificial intelligence, beam diagnostics, control systems, machine learning, particle accelerators, real-time control, reinforcement learning (RL)
in
IEEE Transactions on Nuclear Science
volume
68
issue
8
article number
9442681
pages
7 pages
publisher
IEEE - Institute of Electrical and Electronics Engineers Inc.
external identifiers
  • scopus:85107178866
ISSN
0018-9499
DOI
10.1109/TNS.2021.3084515
language
English
LU publication?
no
additional info
Publisher Copyright: © 1963-2012 IEEE.
id
1326aaba-d139-4f35-b7c6-74ef98579880
date added to LUP
2024-11-26 09:56:50
date last changed
2024-11-26 11:43:54
@article{1326aaba-d139-4f35-b7c6-74ef98579880,
  abstract     = {{<p>Coherent synchrotron radiation (CSR) is generated when the electron bunch length is in the order of the magnitude of the wavelength of the emitted radiation. The self-interaction of short electron bunches with their own electromagnetic fields changes the longitudinal beam dynamics significantly. Above a certain current threshold, the micro-bunching instability develops, characterized by the appearance of distinguishable substructures in the longitudinal phase space of the bunch. To stabilize the CSR emission, a real-time feedback control loop based on reinforcement learning (RL) is proposed. Informed by the available THz diagnostics, the feedback is designed to act on the radio frequency (RF) system of the storage ring to mitigate the micro-bunching dynamics. To satisfy low-latency requirements given by the longitudinal beam dynamics, the RL controller has been implemented on hardware (FPGA). In this article, a real-time feedback loop architecture and its performance is presented and compared with a software implementation using Keras-RL on CPU/GPU. The results obtained with the CSR simulation Inovesa demonstrate that the functionality of both platforms is equivalent. The training performance of the hardware implementation is similar to software solution, while it outperforms the Keras-RL implementation by an order of magnitude. The presented RL hardware controller is considered as an essential platform for the development of intelligent CSR control systems.</p>}},
  author       = {{Wang, Weija and Caselle, Michele and Boltz, Tobias and Blomley, Edmund and Brosi, Miriam and Dritschler, Timo and Ebersoldt, Andreas and Kopmann, Andreas and Garcia, Andrea Santamaria and Schreiber, Patrick and Brundermann, Erik and Weber, Marc and Muller, Anke Susanne and Fang, Yangwang}},
  issn         = {{0018-9499}},
  keywords     = {{Artificial intelligence; beam diagnostics; control systems; machine learning; particle accelerators; real-time control; reinforcement learning (RL)}},
  language     = {{eng}},
  number       = {{8}},
  pages        = {{1794--1800}},
  publisher    = {{IEEE - Institute of Electrical and Electronics Engineers Inc.}},
  series       = {{IEEE Transactions on Nuclear Science}},
  title        = {{Accelerated Deep Reinforcement Learning for Fast Feedback of Beam Dynamics at KARA}},
  url          = {{http://dx.doi.org/10.1109/TNS.2021.3084515}},
  doi          = {{10.1109/TNS.2021.3084515}},
  volume       = {{68}},
  year         = {{2021}},
}