Split step beam propagation method-based underwater propagation modeling and deep-learning phase compensation for optical wireless communication enhancement

Abstract Underwater optical wireless communication suffers from strong turbulence-induced phase distortions, making accurate modeling and compensation critical. In this work, we develop an underwater propagation model based on Split-Step Beam Propagation Method (SSBPM) and evaluate the transmission of Laguerre–Gaussian (LG) beams through modeled turbulence. The model enables analysis of mode-dependent propagation characteristics and indicates improved bit-error-rate (BER) performance for higher-order LG modes. To further mitigate turbulence-induced degradation, we introduce a deep-learning-based phase compensation approach using pix2pix, in which a Fourier Ring Correlation loss is incorporated to preserve high spatial frequency components of the phase. Simulation results show that the proposed data-driven phase compensation suppresses scintillation and enhances communication performance, yielding up to 2.5 dB improvement in carrier-to-noise ratio (CNR) at BER = 10⁻⁹ compared with uncompensated Gaussian beams. The results demonstrate that combining SSBPM modeling with deep-learning-based compensation is effective for improving underwater optical wireless communication.