Convolutional Neural Network Based Intensity-Only Orbital Angular Momentum Mode Decomposition for Free-space Turbulence Compensation

An orbital angular momentum mode decomposition (OAM-MD) technique based on convolutional neural network (CNN) is proposed for the free-space atmospheric turbulence (AT) compensation. The correlation between reconstructed and ground truth patterns can achieve 0.9935 and 0.9808 for 3-OAM model and 5-OAM model, respectively. The proposed turbulence compensation method performs high accuracy in the free-space optical (FSO) link under turbulence strengths D/r 0 = 1, 2, 3 in simulation. The received power of the transmitted beam has 2.5 dB, 7 dB, and 11 dB improvement under the turbulence strength of D/r 0 = 1, 2, and 3 respectively.