Abstract
Offline-trained surrogates for Inertial Confinement Fusion (ICF) suffer a well-known failure mode where iterative optimizers drive inputs into out-of-distribution (OOD) regions, leading to unreliable predictions. Here we present Co4ICF, a co-evolving framework that couples a physics-informed surrogate with a PPO-based pulse optimizer. The surrogate is iteratively fine-tuned on policy-induced trajectories, correcting extrapolation errors as the optimizer shifts the input distribution; the optimizer queries this evolving surrogate as a fast environment. In the 1D MULTI environment, Co4ICF achieves a 146.1% normalized yield based on the current laser design baseline; as a post-hoc cross-fidelity check, the optimized pulse further attains a 246.9% normalized yield when directly evaluated in 2D-MULTI without any 2D training or fine-tuning. Budget-matched ablations support that the gains are not solely explained by additional simulation data and are consistent with the co-evolving mechanism playing a key role. We release a large-scale MULTI-IFE simulation dataset to support future benchmarking.
Blogger's Review: Co4ICF demonstrates immense potential in optimizing inertial confinement fusion by merging physics knowledge with deep reinforcement learning. Its innovative co-evolving mechanism effectively overcomes the limitations of traditional models, laying the groundwork for future applications in more complex environments. This research not only has theoretical significance but also provides strong experimental support for practical applications.