High Performance Signal Design for Optical OFDM Systems using Variational Autoencoder

TL;DR

This paper introduces a variational autoencoder-based method to optimize optical OFDM signals, achieving lower PAPR and improved transmission reliability compared to existing techniques.

Contribution

It presents a novel VAE-based signal design that jointly optimizes constellation geometry and probability for optical OFDM systems.

Findings

Significantly reduces PAPR compared to traditional methods.

Maintains low SER and high mutual information across various SNRs.

Demonstrates superior performance through simulation results.

Abstract

This letter proposes a design of low peak-to-average power ratio (PAPR), low symbol error rate (SER), and high data rate signal for optical orthogonal frequency division multiplexing (OFDM) systems. The proposed design leverages a variational autoencoder (VAE) incorporating gradual loss learning to jointly optimize the geometry and probability of the constellation's symbols. This not only enhances mutual information (MI) but also effectively reduces the PAPR while maintaining a low SER for reliable transmission. We evaluate the performance of the proposed VAE-based design by comparing the MI, SER, and PAPR against existing techniques. Simulation results demonstrate that the proposed method achieves a considerably lower PAPR while maintaining superior SER and MI performance for a wide range of SNRs.