LSTM recurrent neural network prediction algorithm based on Zernike modal coefficients

Abstract

In order to solve the delay problem of the adaptive optics system, an adaptive optics prediction algorithm based on LSTM recurrent neural network is proposed. On this basis, in order to improve the performance of the prediction control algorithm, the relation matrix between driving voltages and Zernike modal coefficients is established. In the prediction model, the Zernike modal coefficients with smaller prediction dimension replace the driving voltages with larger prediction dimension, thus reducing the prediction time. Experimental results show that the proposed algorithm can better correct the wavefront aberration of the beam, and the Strehl Ratio of the beam is significantly improved under different turbulence conditions. Compared with the adaptive optics system which directly predicts the voltages of deformable mirror(DM), this proposed method can effectively shorten the prediction time and has practical value.