Machine learning-based error compensation for high precision laser arbitrary beam splitting

Abstract

A method for error compensation based on machine learning is developed for high-precision laser arbitrary beam-splitting technology. The preliminary approach in laser arbitrary beam splitting is to generate a phase hologram using the iterative Fourier transform algorithm (IFTA) and modulate the incident light beam into multiple beams using a spatial light modulator (SLM). The inherent error in the algorithm and experiment described above prevents beam splitting from improving precision. Error compensated machine learning is developed to mitigate the aforementioned impact and improve beam splitting precision. The corresponding supervised learning regression task on the Numerical simulation dataset, and the SLM experimental dataset establishes two types of mapping relationships between the target image and the detection result image. With the benefit of error compensated machine learning, the mean absolute error (MAE) of beam splitting was reduced by 28% in theory and 21% in the experiment. The error compensated machine learning method is an efficient way to achieve high precision laser arbitrary beam splitting.