Effects of underwater optical turbulence on light carrying orbital angular momentum and its classification using machine learning

Abstract

Laser-based optical links offer the possibility of secure, high data rate underwater communication. Challenges include quickly modulating the light, and rapidly and reliably decoding the received symbols, despite degradation due to optical turbulence and scattering. This paper builds upon a framework introduced in 2016 to address these challenges. We constructed two novel 5-beam Laguerre-Gaussian basis sets, creating 2⁵ symbol alphabets via superposition. We collected 139,000 images of the symbols, transmitted 4.3 meters through water with very strong optical turbulence. We use a novel convolutional neural network to decode them. We believe this work is the first to achieve high accuracy (93.7–99.9%) in an environment with extremely high levels (2∼10⁷-10⁸ m^-2/3) of carefully documented experimentallyinduced optical turbulence. We improve upon previous work in several ways: using a larger alphabet (47 vs. 32), faster classification rates (10,000 Hz vs. 1,000 Hz), and a larger set of test images (96,000 vs 9,600).

KEYWORDS:

• Orbital angular momentum
• optical turbulence
• Laguerre–Gaussian beams
• underwater communications
• convolutional neural networks
• image classification

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

Notes: N/R: not reported.

^a Turbulence given in terms $\frac{D}{r_o}=\frac{\text{receiver diameter}}{\text{Fried coherence length}}$.

Additional information

Funding

This work was funded by U.S. Naval Research Laboratory and Office of Naval Research.