Multi-mission cascaded DNN-based backscattering interference compensation for full-duplex underwater wireless optical...

Area of Science:

• Photonics and Optical Communications
• Machine Learning Applications
• Underwater Exploration Technologies

Background:

• Underwater wireless optical communication (UWOC) offers high data rates but faces challenges with backscattering interference (BI) in full-duplex (FD) systems.
• Existing BI mitigation techniques reduce bandwidth efficiency.
• Sensitive detectors in FD-UWOC systems suffer from nonlinear saturation and signal distortion due to BI.

Purpose of the Study:

• To propose a novel deep neural network for effective BI compensation in FD-UWOC systems.
• To enhance multipath channel impulse response estimation in volumetric scattering environments.
• To address nonlinear saturation effects and signal distortion caused by BI.

Main Methods:

• Development of a multi-mission cascaded deep neural network (MMC-DNN) for BI compensation.
• Implementation of a single-wavelength simultaneous FD-UWOC system.
• Experimental validation in a 1m water tank with realistic volumetric backscattering.

Main Results:

• MMC-DNN significantly reduces Bit Error Rate (BER) from 4.4×10⁻¹ to 7.0×10⁻³ at minimum effective signal optical power (ESOP), improving SNR by 17.8 dB.
• The proposed method outperforms Linear Minimum Mean Square Error (LMMSE), which achieved a BER of 4.1×10⁻² with 13.0 dB SNR improvement.
• MMC-DNN achieves error-free communication at 27.3 dB SNR as ESOP increases, while LMMSE performance plateaus.

Conclusions:

• The MMC-DNN provides superior BI compensation compared to linear methods in FD-UWOC.
• This deep learning approach enhances the feasibility of high-performance underwater communication systems.
• The study demonstrates a significant improvement in communication reliability for offshore and marine applications.