Machine Learning-based Design of Structured Laser Light for Improved Data Transfer Rate in Underwater Wireless Communication
Abstract
A system using Laguerre-Gaussian (LG) beams of structured light and deep convolutional neural networks (CNNs), is utilized. The structured beams of light are encoded to carry information, with each combination resulting in a distinct image. This creates an alphabet of 2N, images, where N is the number of basis beams and bits encoded per message. For this investigation, a novel methodology for optimizing network alphabet design is proposed, and 256 and 1024-beam alphabets are designed for optimal classification accuracy through optical turbulence using a CNN. For simulated evaluation, the beams were propagated using the split-step method through random phase screens drawn from the Nikishov spectrum for oceanic turbulence. In the experimental environment, the beams were propagated over ~2.5 meters through optically turbulent water with strong turbulent fluctuations. This study is novel in its use of the scintillation of a Gaussian beam to estimate the strength of the turbulence in real-time. In the simulated environment, we report 100 classification accuracy for the 256-beam alphabet, indicating the CNNs ability to learn weak fluctuations. Under experimental conditions, we report over 97 percent accuracy for 256-beam alphabets, and over 90 accuracy for the 1024-beam alphabet.
Document Details
- Document Type
- Technical Report
- Publication Date
- May 16, 2022
- Accession Number
- AD1171853
Entities
People
- William A. Jarrett
Organizations
- United States Naval Academy