Author: Tomin, S.
Paper Title Page
THPAB191 Physics-Enhanced Reinforcement Learning for Optimal Control 4150
 
  • A.N. Ivanov, I.V. Agapov, A. Eichler, S. Tomin
    DESY, Hamburg, Germany
 
  We propose an approach for incorporating accelerator physics models into reinforcement learning agents. The proposed approach is based on the Taylor mapping technique for simulation of the particle dynamics. The resulting computational graph is represented as a polynomial neural network and embedded into the traditional reinforcement learning agents. The application of the model is demonstrated in a nonlinear simulation model of beam transmission. The comparison of the approach with the traditional numerical optimization as well as neural networks based agents demonstrates better convergence of the proposed technique.  
poster icon Poster THPAB191 [0.846 MB]  
DOI • reference for this paper ※ https://doi.org/10.18429/JACoW-IPAC2021-THPAB191  
About • paper received ※ 11 May 2021       paper accepted ※ 29 July 2021       issue date ※ 24 August 2021  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)