Orbital angular momentum distortion correction and mode identification using a multitask neural network

Vortex beams with orbital angular momentum (OAM) can meet the demands of high-capacity modern communication and significantly increase the transmission capacity of underwater wireless optical communication (UWOC) systems. However, the beam distortion caused by oceanic turbulence (OT) poses challenges for OAM transmission and identification. In this paper, we address these problems by proposing a technique based on a multitask neural network (MTNN) that can achieve high-quality distortion correction, identify OAM modes, and resist or eliminate OT. The MTNN model performs feature extraction on the distorted OAM intensity distribution, and the two output branches utilize the shared feature map to output the predicted OT phase screen and OAM mode. The results show that the MTNN model in the proposed scheme can eliminate OT in the channel, correct the OAM distortion, and accurately identify its mode. The mode purity and identification accuracy of the OAM after distortion correction are significantly improved. Compared with the common convolutional neural network (CNN) model, the MTNN model demonstrates superior antiturbulence performance at different OT strengths. The proposed scheme provides new technologies for the innovative development of high-performance UWOC systems.