Efficient synchronous retrieval of OAM modes and AT strength using multi-task neural networks

Area of Science:

• Optical Communications
• Atmospheric Physics
• Machine Learning

Background:

• Orbital Angular Momentum (OAM) beams offer high capacity for optical communication.
• Atmospheric turbulence causes OAM beam distortion, leading to signal interference and reduced communication quality.
• Stable OAM mode recognition is crucial for reliable free-space optical communication.

Purpose of the Study:

• To develop a stable and efficient method for simultaneous recognition of OAM modes and atmospheric turbulence intensity.
• To enhance the robustness of free-space optical communication systems against atmospheric disturbances.

Main Methods:

• Established an equivalence between a continuous dynamics system and a stable network unit (RUEM).
• Proposed a multitask neural network, OATNN, incorporating RUEM for simultaneous OAM mode and turbulence intensity recognition.
• Conducted numerical experiments to validate the model's stability and accuracy.

Main Results:

• OATNN achieved 99.37% accuracy in recognizing four turbulence intensity levels.
• The network demonstrated 99.05% accuracy in recognizing ten OAM modes.
• Performance was evaluated in a 2000m channel transmission scenario.

Conclusions:

• The proposed OATNN model, embedded with RUEM, offers a stable and accurate solution for OAM mode and turbulence intensity recognition.
• This approach significantly improves the quality and reliability of free-space optical communication in turbulent atmospheric channels.