Numerical Simulation of a Multilayer Perceptron Neural Network for Predicting and Correcting the Wavefront in an Adaptive Optics System for Closed and Open Loops

Abstract

This paper presents a numerical simulation of a multilayer perceptron neural network (MLPNN) for predicting and correcting the disturbed wavefront in both closed and open loops in adaptive optics systems. Traditional adaptive optics systems often encounter limitations restraining their efficiency. The MLPNN model was implemented to reduce the challenges by enhancing prediction and correction accuracy in both closed and open loops. It was trained and tested over a wide range of turbulence marked by Fried parameters  . Under weak turbulence,  closed loop achieved a Strehl ratio (SR) up to 0.9283 and the average RMSE (ARMSE) of 0.2857 compared to 0.8414 for SR and 0.4411 ARMSE for open loop. In strong turbulence,  SR was 0.4328 and 0.9394 for ARMSE in closed loop, whilst SR degraded to 0.1467 with a rise in ARMSE to 1.3732 in open loop. Compared to the traditional method, SR improved by up to 38% and reduced ARMSE by up to 99%.  These results confirm that closed loop consistently outperformed in both high and low turbulence conditions using the MLPNN approach.