Learning diffractive optical communication around arbitrary opaque occlusions

Area of Science:

• Optics and Photonics
• Communication Systems
• Artificial Intelligence

Background:

• Free-space optical communication (FSOC) is hindered by occlusions blocking the light path.
• Maintaining communication links in the presence of arbitrary, opaque obstacles is a significant challenge.

Purpose of the Study:

• To demonstrate a direct communication scheme that can circumvent fully opaque, arbitrarily shaped occlusions in free-space optical links.
• To develop a robust method for transmitting optical information around line-of-sight blockages.

Main Methods:

• A deep learning approach was used to jointly train an electronic neural network encoder and a passive, all-optical diffractive network decoder.
• The system was designed to transfer optical information around occlusions of arbitrary shapes.
• Experimental validation was performed in the terahertz spectrum using a 3D-printed diffractive decoder.

Main Results:

• The trained encoder-decoder pair successfully transmitted arbitrary optical information around opaque occlusions.
• Information decoding occurred at the speed of light via passive light-matter interactions.
• The system demonstrated resilience to unknown changes in occlusion shape and size.

Conclusions:

• This novel scheme enables free-space optical communication around opaque obstacles, overcoming a major limitation.
• The demonstrated method is scalable to various wavelengths and offers high data-rate potential for future communication systems.
• The use of a jointly trained neural network and diffractive decoder provides a robust and efficient solution for non-line-of-sight optical communication.