Speaker
Description
Robust accelerator control increasingly relies on data-driven optimisation, yet balancing adaptability with safety remains challenging. Simulation-driven physics-informed reinforcement learning (RL) relies on soft constraints without firm safety guarantees, and classical matrix inversion becomes suboptimal under noise and hard actuator limits. Using the AWAKE electron beam steering task at CERN as a high-fidelity benchmark, we formulate beam steering as a stochastic control problem in a linear Markov Decision Process with continuous state and action spaces and realistic constraints, and compare classical inversion, Model Predictive Control (MPC), data-driven Gaussian-Process MPC (GP-MPC) and RL.
Our main contribution is a Causal GP-MPC scheme that embeds the beamline’s causal layout directly into the GP prior and kernel design. This structural inductive bias reduces model complexity, improves conditioning, and enables accurate multi-step prediction from limited data. In simulation studies based on the measured response matrix, Causal GP-MPC achieves performance comparable to MPC with the perfect model while requiring only observational data. It outperforms unstructured GP-MPC and RL baselines in sample efficiency, noise robustness, and online optimisation time. Taken together, these results demonstrate that causally structured learning offers a promising pathway toward data-efficient, interpretable, and deployable control strategies for complex accelerator systems.
| In which format do you inted to submit your paper? | LaTeX |
|---|---|
| Preprint marking on your proceeding paper | I wish my paper to be marked as preprint. |