REINFORCEMENT LEARNING APPROACHES FOR BEAM DYNAMICS OPTIMIZATION IN PARTICLE ACCELERATORS

Particle accelerators are pivotal instruments in both fundamental research and practical applications across various scientific, technical, and industrial fields. These machines are composed of different complex sections that allow researchers to study the particles. In particular, this thesis aims to utilize Reinforcement Learning algorithms to build a model that can tune accelerator's parameters with the objective of optimizing the task of beam emittance optimization. This analysis is conducted using the particle accelerators at the INFN National Laboratories of Legnaro in Italy and the results show how the application of Reinforcement Learning can improve the optimization despite changes in parameters. The results obtained can be applied to other facilities that face similar challenges.