A Meta-Learning-Enhanced Deep Learning Approach for Underwater Wireless Optical Communication Channel Emulator

The complexity and variability of marine environments pose significant challenges to modeling underwater wireless optical communication (UWOC) systems. Most conventional modeling techniques are tailored to specific datasets from particular water bodies, limiting their ability to generalize across new tasks. This study proposes and validates a UWOC channel emulator based on a meta-learning strategy combined with a deep convolutional conditional generative adversarial network (DCGAN) to address these challenges. The emulator efficiently learns the characteristics of various channel responses and generates simulated signals in novel environments with specific channel conditions. These signals closely match the statistical properties of real UWOC systems. Compared to methods based on multilayer perceptron, convolutional neural networks, and DCGAN, the experimental results demonstrate that the proposed method achieves superior performance in both the time and frequency domains across underwater environments with varying attenuation levels. This underscores its robust generalization capability and establishes a promising direction for applying deep learning methods to UWOC channel modeling.