Deep learning-based robust OAM mode recognition in atmospheric turbulence

Vortex beams carrying orbital angular momentum (OAM) offer a promising approach to increasing channel capacity in free-space optical (FSO) communication by enabling multiplexing through topological charge (TC) states. However, atmospheric turbulence introduces significant distortions, posing challenges for accurate OAM recognition. In this paper, we propose ARNet, a deep-learning-based network that combines channel-attention-enhanced residual blocks with dynamic L2 regularization to enhance the robustness of OAM identification under turbulence. Experiments demonstrate that ARNet achieves 95.52% recognition accuracy for OAM states ranging from -5 to +5 under strong turbulence (Cn2=5×10^-13m^-2/3) conditions. The proposed method significantly outperforms traditional interference-based techniques, demonstrating enhanced resilience and potential for practical FSO system deployment in turbulent environments.