Classifying beams carrying orbital angular momentum with machine learning: tutorial.

Area of Science:

• Optics and Photonics
• Machine Learning
• Optical Communications

Background:

• Orbital angular momentum (OAM) offers novel ways to encode information in light.
• Propagating OAM beams through turbulent media causes signal distortion.
• Machine learning (ML) shows promise in analyzing complex optical data.

Purpose of the Study:

• To bridge the knowledge gap between optics and machine learning for OAM communication systems.
• To provide a practical guide for implementing ML in optical communication research.
• To facilitate the adoption of AI in analyzing distorted optical signals.

Main Methods:

• Review of ML applications in optics, focusing on OAM beams.
• Discussion of experimental design for OAM optical communication (Laguerre-Gaussian beams, turbulence generation, image capture).
• Introduction to convolutional neural networks (CNNs) for image classification.

Main Results:

• Demonstration of ML, specifically CNNs, for classifying distorted OAM symbols.
• Provision of experimental design considerations for robust optical communication.
• Establishment of best practices for the field.

Conclusions:

• ML techniques, particularly CNNs, are effective for decoding OAM signals in turbulent environments.
• Accessible resources, including code and a dataset, are provided for researchers.
• This work encourages interdisciplinary research in AI-powered optical communications.