Embedding AI into laser pulse shaping closed-loop control

Laser pulse shaping is an essential task for power balance in the high-power laser facility. The precisely shaped laser pulse relies heavily on the waveform generation of arbitrary waveform generator (AWG), which provides the electric pulse drive. Based on a set of control model and iterative algorithm, we established a closed-loop control method for pulse shaping. However, it consumes too much time, greatly affecting the working efficiency of laser physics experiments. To improve the time efficiency of closed-loop control, we propose an artificial intelligence assisted method. By training a U-net model to predict the initial pulse waveform, the closed-loop control process of pulse shaping is greatly accelerated. Furthermore, given that the artificial intelligence model is not compatible with the current pulse shaping control platform, we introduce thrift as a fresh middle ware, and implement an RPC bridge. Thus the artificial intelligence model is smoothly embedded into the traditional closed-loop control process.

Automated summary

Laser pulse shaping is crucial for power balance in high-power laser facilities. We propose an artificial intelligence assisted method to accelerate the closed-loop control process by training a U-net model to predict the initial pulse waveform.