Real-time Artificial Intelligence for Accelerator Control: A Study at the Fermilab Booster

TL;DR

This paper presents a neural network-based reinforcement learning approach for real-time regulation of accelerator magnet power supplies, demonstrating the potential for improved operational stability at Fermilab's Booster.

Contribution

It introduces a novel method using surrogate models and neural networks trained via reinforcement learning for precise accelerator control, including deployment on FPGA hardware.

Findings

Preliminary results show effective regulation using surrogate models.

Neural networks can be compiled for FPGA deployment.

Potential for enhanced stability in accelerator operations.

Abstract

We describe a method for precisely regulating the gradient magnet power supply at the Fermilab Booster accelerator complex using a neural network trained via reinforcement learning. We demonstrate preliminary results by training a surrogate machine-learning model on real accelerator data to emulate the Booster environment, and using this surrogate model in turn to train the neural network for its regulation task. We additionally show how the neural networks to be deployed for control purposes may be compiled to execute on field-programmable gate arrays. This capability is important for operational stability in complicated environments such as an accelerator facility.