Speaker
Description
The Electron-Ion Collider (EIC) will rely on unprecedented levels of machine-assisted optimization, predictive modeling, and autonomous control to achieve its performance goals. To prepare for this challenge, the multi-laboratory EIC-BeamAI collaboration has been established and funded to develop and deploy advanced machine learning and digital-twin technologies within the EIC pre-accelerator chain at Brookhaven National Laboratory. Working with Linac, Booster, AGS, and the NSRL line, and associated beamlines, the collaboration is gathering operational experience in ML-based optimization of 3D beam emittances and beam polarization, and in real-time, AI-assisted decision making. These testbeds allow us to probe the integration of differentiable simulations with SciBmad, Bayesian optimization, and reinforcement learning approaches with actual accelerator operation. The resulting insights will form the foundation for robust digital twins, streamlined commissioning strategies, and reliable ML-enabled control of the full EIC. This talk will present the structure and goals of the EIC BeamAI collaboration, early results from pre-accelerator deployments, and a roadmap toward ML-empowered EIC operations.
Funding Agency
DOE-AP through grants DE-SC0025351 and DE-SC0024287
| In which format do you inted to submit your paper? | LaTeX |
|---|