Experimental Assessment of the Atmospheric Effects on Laser Communications on Maritime Environment

Abstract

This dissertation investigates the free space optical communications (FSOCs) technology, the so-called laser communications (LaserComm). The FSO technology operates in the optical and infrared spectrum for data transmission and offers significant advantages comparing to their radio frequency (RF) counterparts. These include increased data rate transmission, greater security and immunity, lower cost of installation and, finally, no license restrictions. The major challenge that FSO systems face, is the degradation effects of the atmosphere and in especially the optical turbulence. FSO technology applications span from terrestrial and underwater to space. Each FSO application face different kind of challenges due to the variety of the atmospheric conditions that a propagating beam encounters. An FSO link over a maritime environment is investigated. The optical turbulence phenomenon in a maritime environment exhibits significant differences with a terrestrial one and causes the effect of scintillation, that is the spatial and temporal variation of the irradiance in the receiver. To study these variations, statistical distributions are required. Already, many different probability density functions (PDFs) have been studied and related the received irradiance with the optical turbulence effect, quantified by the refractive index structure parameter, Cn^2. The major goal of this dissertation is the experimental assessment of the optical turbulence effect towards an FSO link over maritime environment. We utilized the experimental site of the Hellenic Naval Academy (HNA), Piraeus, Greece as well as the site of the Naval Postgraduate School (NPS) in Monterey, California to collect macroscopic meteorological parameters (wind speed, relative humidity, air temperature, solar radiation, dew point, air pressure and rainfall rate), Cn^2 values with different methods and the received signal strength (RSSI) on the FSO terminal. Based on this data, initially we developed second-order polynomial models for the RSSI parameter, using regression modeling, to quantify its relation with the macroscopic environmental parameters and demonstrated the prediction accuracy of various machine learning based algorithms (Random Forest, Gradient Boosting, Decision Trees, k-Nearest Neighbors, Neural Networks) for an FSO link performance. The similarity between different probability distributions is investigated, by employing an information theoretic method, namely the so-called Jensen–Shannon divergence, a symmetrization of the Kullback–Leibler divergence. The outage probability of the link is estimated by utilizing a theoretical model and the operational status of the link, either functional or not, is modeled by a classification deep neural network (DNN). Finally, the refractive index structure parameter, Cn^2, is modelled using ML-based regression (Random Forest, ANN και Gradient Boosting) and classification algorithms (DNN).