Deep Learning Aided Misalignment-Robust Blind Receiver for Underwater Optical Communication

Abstract

Underwater wireless optical communication (UWOC) has been proposed to provide high-rate data services by exploiting the ample optical spectra. However, the underwater scenario presents a hostile environment for wireless optical signal propagation due to the various channel effects, such as absorption, scattering and turbulence. Furthermore, link misalignment (LM) between the optical transmitter and receiver caused by the turbulent water waves degrades the achievable system performance. All the aforementioned factors make the information recovery a challenging task for UWOC systems, especially for long-distance data transmissions. In this letter, we introduce a deep learning (DL) based misalignment-robust blind receiver (MBR) to recover the received data in a multiple-input multiple-output (MIMO) UWOC system, where a convolutional neural network (CNN) is used to formulate the signal characteristics in model training, a CNN combiner is utilized for characteristic analysis and combination, and a CNN demodulator is applied to recover the transmitted information. Evaluation results demonstrate that a reliable performance is achievable by the proposed DL-MBR scheme in UWOC scenarios when a relatively large LM occurs.