Optimized attention neural network based joint link misalignment and beam divergence power optimization for underwater optical wireless communications
K. Bala,D. Venu,R. Prakash Kumar,K. Ashok Kumar,S. Varalakshmi,Guda Chenna Reddy
DOI: https://doi.org/10.1007/s11082-024-07054-7
IF: 3
2024-05-24
Optical and Quantum Electronics
Abstract:Underwater Wireless Optical Communication (UWOC) offers a promising alternative to acoustic transmission methods, boasting reduced latency, increased data rates and smaller device sizes, with a focus on leveraging the underwater channel's optical spectra diversity for high-speed data services. Underwater signal propagation encounters absorption, turbulence and scattering, causing misalignment between optical transmitter and receiver due to turbulent water waves, degrading performance. In Point-to-Point UWOC, changes in beam divergence angle at the transmitting node disrupt links between underwater sensor nodes and maximize sensor battery usage. Therefore, this research work introduces the Efficient Deep Spiking Convolutional Neural Network with Graph Attention Layer (EDSCNN-GAL) that serves as a misalignment-robust blind receiver, utilizing convolutional layers, spatial-pooling layers and graph attention layers to extract and process features from received optical signals. Additionally, Multi scale Deep CapsNet (MD CapsNet) optimizes transmission power and beam divergence angle to sustain connections and minimize battery usage in underwater sensors. Enhanced walrus optimization is employed to efficiently optimize MD CapsNet parameters. The introduced model's efficacy is demonstrated through a comparison with state-of-the-art models, achieving better, throughput superior instantaneous signal to noise ratio (24 dB), lower misalignment probability (4%) and less transmission power (6.4 mW).
engineering, electrical & electronic,optics,quantum science & technology