Bayesian and particle swarm approaches to inertial confinement fusion optimization

The optimization of laser pulse shapes and target configurations is central to high-performance inertial confinement fusion (ICF) implosions, yet remains challenging due to the high dimensionality of the design space and the substantial computational and experimental cost of evaluation. This work presents, to our knowledge, the first comparison of Gaussian process-based Bayesian optimization and particle swarm optimization (PSO) frameworks augmented with physics-motivated extensions in full radiation-hydrodynamic ICF optimization under experimentally relevant constraint handling. These methods are first applied to the re-optimization of low-convergence-ratio wetted-foam implosions, providing a benchmark against traditional sequential scan approaches. Both strategies identify improved designs, with Bayesian optimization achieving the highest final performance using fewer simulations, while PSO converges more rapidly in wall-clock time. The PSO framework is then extended to a 16-dimensional fast ignition design problem, where Gaussian process-based Bayesian optimization becomes computationally impractical. In this regime, PSO efficiently identifies a compressed fuel assembly with ρR≈1.5 g/cm2 under strict laser intensity and energy constraints. These results demonstrate that the presented optimization strategies outperform conventional scan-based approaches and provide a scalable platform for high-dimensional ICF optimization. Beyond numerical design studies, the same frameworks are directly applicable to experimental optimization campaigns on high-power laser facilities, where limited shot availability and high evaluation cost demand efficient search methodologies.