The workshop will focus on addressing critical challenges in fusion simulation applied for the design of optimal plasma experiments and designing plasma controllers, which are essential for advancing fusion energy. Recent breakthroughs, such as the ignition success at NIF, highlight the crucial role of computational and data-driven models in guiding experimental setups. Future progress will depend on developing high-fidelity computational models informed and validated by experimental data, and reduced-order and surrogate models for many-query applications like optimization and control. These advances will help bring sustainable fusion energy closer to reality by optimizing experiment designs and interventions.
To achieve these goals, the workshop will bring together experts from plasma physics, computational mathematics, and statistics to tackle the complexities of large-scale optimization and decision-making in plasma experiments. Discussions will explore the use of surrogate models, machine learning, and multi-fidelity methods to address challenges like model complexity, computational expense, and uncertainty. Emphasis will be placed on developing scalable algorithms for optimal control and design under uncertainty, with a focus on leveraging both high-fidelity and low-fidelity models. The workshop will also explore reinforcement learning and inverse problem-solving methods to guide experimental data acquisition and improve predictive accuracy in plasma physics applications.
Jonathan Citrin
(DeepMind Technologies)
Ana Kupresanin
(Lawrence Berkeley National Laboratory)
Cristina Rea
(Massachusetts Institute of Technology)
Tim Wildey
(Sandia National Laboratories)
Copyright. All Rights Reserved.