Machine learning for robust classical and quantum communications
Abstract
The primary objective of this project is to develop innovative techniques that will advance the frontiers of communications. This involves the development, optimization, and experimental verification of a novel suite of neural networks to substantially enhance the robust implementation of free-space optical (FSO) communications. The methods developed will be directly applicable to next-generation quantum communications systems, with many tools applicable to imaging schemes. This machine learning toolbox consists of a neural network feedback system to correct for the negative effects of turbulent propagation on optical states, a neural network for symbol and image discrimination applicable to numerous keying and modulation schemes, and a neural network for discovering optimal arbitrary communications platforms. We will apply these neural network systems to a wide range of communications scenarios, to optimize and stabilize the signal intensity at the fiber coupler of a receiver, to predict and correct for atmospheric fluctuations of attenuation, and to correct for various temporal effects in time-bin type keying systems. The results of the project will directly address the Navy~s vision ~to provide high throughput robustcommunications and networking to ensure all warfighters ~ from the operational command to the tactical edge ~ have access to information, knowledge, and decision-making necessary to perform their assigned tasks,~ by enhancing the implementation of ~reliable, interoperable, survivable, secure, and timely communications and networking, and the availability of high capacity multimedia (voice, data, imagery) communication networks [that] is fundamental to nearly allDepartment of Navy missions.~The results of the project will have significant impacts on the ONR and multiple DoD agencies, by addressing many problems encountered in free-space optical communications in a novel manner. The results will be directly applicable to current FSO communications systems, including OOK, BSPK, and QPSK modulation. Current FSO networks will become more robust and reliable in contested environments and in turbulent media, resulting in enhanced high data rate communications systems. This includes reaching theoretical performance limits in such communications links despite real-world issues encountered, including imperfect detection systems. The results will directly add resiliency and additional capabilities to existing communications systems by increasing signal-to-noise ratio stability, reducing the effects of atmospheric attenuation, lowering bit-error rates (BERs) while increasing bit transmission rates, increasing and stabilizing fiber coupling efficiencies, and correcting for dispersive and temporal effects on optical pulses. The BER of FSO communications systems will be significantly enhanced by incorporating the developed symbol discrimination neural network. Novel and previously undiscovered communications architectures will also be discovered that provide optimal functionality in any given arbitrary communications scenario. Additionally, the results will be directly integrable into current systems, and will allow for high-speed functionality by pre-training the various neural networks involved. Much of the work will also be directly applicable to imaging systems, enhancing the warfighter~s ability to actively extract useful image information, and allow for the robust transmission and reception of images through strongly turbulent media. Lastly, all of the results are effectively future-proof, as they will be easily transitioned into future quantum communications and imaging schemes. The results will also significantly advance our ability to implement quantum-enhanced schemes such as mathematically-secure quantum key distribution,by allowing for the robust propagation of delicate quantum states of light through turbulent media.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- May 23, 2019
- Source ID
- N000141912374
Entities
People
- Ryan T. Glasser
Organizations
- Office of Naval Research
- Tulane University of Louisiana
- United States Navy