Quantum-Classical Autoencoder Architectures for End-to-End Radio Communication

TL;DR

This paper explores hybrid quantum-classical autoencoder architectures for end-to-end radio communication, demonstrating robustness under noisy conditions and providing a framework for future quantum machine learning applications in wireless systems.

Contribution

It introduces a comprehensive framework for hybrid quantum-classical autoencoders in radio communication, including formulas, simulations, and analysis of robustness under noise.

Findings

Models are robust against Gaussian noise and Rayleigh fading.

Hybrid architectures perform well with 4-QAM and 16-QAM schemes.

The framework supports future development of quantum neural networks for communication.

Abstract

This paper presents a comprehensive study on the possible hybrid quantum-classical autoencoder architectures for end-to-end radio communication against noisy channel conditions using standard encoded radio signals. The hybrid scenarios include single-sided, i.e., quantum encoder (transmitter) or quantum decoder (receiver), as well as fully quantum channel autoencoder (transmitter-receiver) systems. We provide detailed formulas for each scenario and validate our model through an extensive set of simulations. Our results demonstrate model robustness and adaptability. Supporting experiments are conducted utilizing 4-QAM and 16-QAM schemes and we expect that the model is adaptable to more general encoding schemes. We explore model performance against both additive white Gaussian noise and Rayleigh fading models. Our findings highlight the importance of designing efficient quantum neural network architectures for meeting application performance constraints -- including data re-uploading methods, encoding schemes, and core layer structures. By offering a general framework, this work paves the way for further exploration and development of quantum machine learning applications in radio communication.