Three novel efficient Deep Learning-based approaches for compensating atmospheric turbulence in FSO communication system

TL;DR

This paper introduces three innovative deep learning-based methods for mitigating atmospheric turbulence in FSO communication, demonstrating comparable performance to traditional systems with reduced complexity across various MIMO configurations.

Contribution

It is the first to apply deep learning, including transceiver and transmitter learning, in FSO-MIMO communication for turbulence compensation.

Findings

Proposed methods achieve similar performance to conventional systems.

Effective across a wide range of turbulence conditions.

Maintain performance with different MIMO configurations.

Abstract

One of the main problems encountered with Free Space Optical (FSO) Communication system is the atmospheric turbulence. Although many solutions exist for combating this effect, they have either high complexity or low performance. In this paper, a comprehensive investigation is developed, and three new effective Deep Learning (DL) based solutions are proposed. This paper, for the first time, deploys Deep Learning, transceiver learning, as well as transmitter learning in FSO communication. In addition, this is the first time that DL approach is implemented in FSO-Multi-Input Multi-Output (MIMO) communication. Results of the proposed structure are compared with the state of the art MQAM based FSO system with Maximum Likelihood Detection. Wide range of atmospheric turbulence, from weak to the strong regime, are considered; results indicate that the proposed structures despite less complexity, have the same performance as the outstanding conventional structure. In addition, in different MIMO structures (different combiners and the different number of transceiver apertures), the proposed structure still achieve the performance of the state of the art conventional system.