Artificial Intelligence for Accelerator Control

TRIUMF accelerator complex enables world class research in fields ranging from nuclear physics to material sciences, life sciences and nuclear medicine offering a major advantage in probing nuclear physics processes and uniquely place TRIUMF as a leader in experiments at the “precision frontier” to address some of the fundamental scientific questions of our time.
To deliver isotope beam produced by the ISAC facility and ARIEL in the future, with the desired intensity and quality to various nuclear physics experiments, multiple parameters of the accelerator need to be tuned during a lengthy, manual procedure. Tuning process must be performed after any adjustments or maintenance of the accelerator. During the course of the project the visiting researcher will develop automated control of a beam line section of the ISAC accelerator utilizing deep reinforcement learning techniques. Simulation environment of the beam line section will be developed for trials of the techniques. Provided successful automated tuning is demonstrated in simulated environment interface to existing accelerator will be developed and automated tuning trialled.

Faculty Supervisor:

Oliver Kester

Student:

Partner:

Rheinisch-Westfälische Technische Hochschule Aachen

Discipline:

Physics

Sector:

Education

University:

TRIUMF

Program: