Residual network-based aberration correction in a sensor-less adaptive optics system

Abstract

The performance of free-space optical communication (FSOC) is often affected by atmospheric turbulence. The sensor-less adaptive optics (SLAO) system is an effective method for overcoming the effects of atmospheric turbulence. The performance of the control algorithm in the SLAO system directly determines whether the SLAO system can effectively correct wavefront aberrations. In this study, we introduce a residual network (ResNet) as a control algorithm to replace the traditional control algorithm. By lowering the number of iterations, this strategy enhances the real-time performance of the FSOC system. The final ResNet model can achieve an accuracy of 0.98 for training and 0.92 for testing. The simulation results show that stochastic parallel gradient descent (SPGD) algorithm takes 700 times longer and requires at least 500 iterations to achieve the same performance as ResNet. And we verify the feasibility of the ResNet model by setting up an experiment.