Autonomous Laser Pulse Optimization on the HILL Laser for High-Repetition-Rate Applications

Recent advances in machine learning (ML) and automated control are enabling significant improvements in the optimization and stability of high-repetition-rate (HRR) laser systems. In this work, we present an autonomous pulse optimization system for the HILL Laser at LLNL, an HRR facility, integrating distributed diagnostics and real-time feedback. Data from key diagnostics—including frequency-resolved optical gating (FROG), energy meter, and spectrometer—are automatically collected and transferred across multiple control computers using the EPICS framework. Traditionally, FROG traces present a bottleneck for accurate real-time analysis; here, we overcome this by training a neural network to rapidly and accurately extract pulse shape, spectrum, and phase from raw FROG data within 50 ms. The extracted parameters are used to drive automated adjustments of the DAZZLER, enabling dynamic compensation of dispersion and rapid convergence to user-defined pulse characteristics. All diagnostics and control loops are fully automated, allowing for continuous optimization and robust performance at a 1Hz repetition rate. This approach demonstrates a scalable pathway for data-driven, autonomous operation of HRR laser facilities for advanced scientific applications.