The Study Of Lognormal-Rician Channel Model For Free Space Optical Communication

The free space optical communication (FSO) has gained rapid development in recent years because of its advantages, including large transmission capacity, rapid transmission speed and rich spectrum resources. The FSO system uses atmospheric channel as its transmission medium. Thus the communication quality depends heavily on channel conditions. Optical beam drift and scintillation, especially the latter, can degrade the performance of the FSO system. They are both caused by atmospheric turbulence. Several methods have been proposed to reduce the influence of the phenomenon, but they all have corresponding limitations.We can consider studying the atmospheric channel and using reasonable channel model to improve the performance of the FSO system. The probability density functions (PDF) of optical intensity are often used to describe scintillation. Among all known models, the Lognormal-Rician (LR) model can fit perfectly with simulation results and experimental data. We can obtain the parameters of the atmospheric channel and know the communication quality of the FSO system in real-time through the study of the LR model parameters. We can switch the optical link when the link is not stable. This strategy can obviously increase the stability of the FSO system.However, the PDF of the LR model has an integral form. It is very complicated to get its closed-form solution. Researches on the LR model are still starting out. In this thesis, we mainly studied the LR model. The main work was as follows:1. We derived and analyzed the PDF of the LR model. Two shape parameters r and σ2 were found. We then used the classical moment estimation method to estimate the two parameters and pointed out shortages of the method.2. We used generalized method of moments (GMM) to do parameter estimations for the LR model. Monte Carlo simulation was used to analyze the asymptotic variance, the mean squared error (MSE) and the normalized mean squared error (NMSE) of estimators.3. We further adapted the expectation maximization (EM) algorithm to do parameter estimations for the LR model. The Cramer-Rao Lower Bound (CRLB) of this method for the estimation of the LR model was derived for the first time. We used the same simulation method to analyze CRLB, MSE and NMSE of estimators.