1st prize at the Reinforcement Learning challenge of the Reinforcement Learning for Autonomous Accelerators workshop

Abstract

The challenge took place during the Reinforcement Learning for Autonomous Accelerators workshop (RL4AA) in Liverpool, UK. It consisted on centering and focusing a particle beam on a diagnostic screen by automatically tuning magnet strengths using Reinforcement Learning (PPO or SAC). This problem is based on a simulated segment of the CLARA particle accelerator at the Cockcroft Institute (UK), where efficiency and precision are key. Thanks to my team Georg Schäfer (FH Salzburg), Filip Peczek (University of Manchester), and Lana Buckleton (University of Strathclyde) who truly made this possible.