| Vortex beams carrying orbital angular momentum greatly increase the capacity and efficiency of optical communications due to their orthogonal characteristics,and they have excellent safety and flexibility,which have broad application prospects in optical communications.When a vortex beam propagates in free space,it is inevitably affected by atmospheric turbulence,causing random fluctuations in phase amplitude and distortion of the wavefront,resulting in degradation of the performance of the entire vortex optical communication system.One of the effective solutions is to introduce adaptive optics to improve the performance of the system by correcting the wavefront aberrations.The classical wavefront compensation method is inadequate as the application scenario becomes more complex,and this paper conducts a study to improve the compensation method.The original non-convex optimisation algorithm is improved upon to improve the compensation for beam phase distortion due to atmospheric turbulence effects and pointing error effects in vortex optical communications.The improved non-convex optimisation algorithm uses the robustness of the sample median in the initialisation calculation to formulate new truncation rules to reduce the effect of too large and too small outliers on the results through the median value.In the iterative optimization calculation stage,in order to make the iteration direction always close to the true value,the adjacent iteration results are reweighted once in the calculation,and finally the new weight coefficients are applied in the calculation of gradient descent to improve the compensation accuracy of the algorithm.The non-convex optimization algorithm based on median reweighting achieves better results than the original algorithm in experimental simulations in different channel environments and different modes,with the recovered Strehl ratio reaching 0.9 in the case of medium turbulence and 0.81 with pointing error.In the OAM multiplexing transmission detection experiments,the BER gain of the system compensated by the improved algorithm is more than 1d B.A wavefront aberration compensation method based on a residual attention neural network is proposed to address the problems of phase aberrations caused by atmospheric turbulence in vortex beams,the limited performance of traditional iterative optimisation algorithms and the insufficient ability of deep convolutional neural networks to extract spot features.The overall structure of the network is a residual network,which uses multi-scale operations for distributed extraction of light spot features and increases the perceptual field information of the feature map.By embedding a hybrid attention mechanism of channel attention and spatial attention in the network,the network’s ability to extract aberrated spot features is improved.The overall network is improved on the original residual network,which greatly improves the network’s ability to map aberrated light intensity images to Zernike coefficients.The loss function is designed in conjunction with realistic evaluation metrics in the network training,thus increasing the realism of the output Zernike coefficients.The wavefront compensation method based on the residual attention network is experimentally simulated in different channel environments and gives better results than the previous network model.The improved model has a BER gain of about 0.6 d B with PV=0.045 rad and RMS=0.011 rad for weak turbulence,1 d B BER gain with PV=0.226 rad and RMS=0.046 rad for medium turbulence and 1.1 d B BER gain with PV=0.275 rad and RMS=0.071 rad for strong turbulence.In the simulations under different turbulence conditions,the improved network recovered the smallest peak-to-trough and root-meansquare values,and its residual phase was closest to the plane and had the lowest BER in the multiplexed communication system,which supports a wider application of deep learning in wavefront compensation. |
