Echo State Network for Turbulence-Induced Fading Channel Prediction in Free-Space Optical Systems

Free-space optical (FSO) communication has established its reputation for delivering ultra-high data-rate services. Over the time-varying turbulence channels, the channel state information (CSI) tends to be outdated due to the feedback delay. This may lead to considerable performance deterioration in FSO systems. This paper studies a prediction scheme for accurately updating CSI over FSO turbulence-induced fading channels. We employ the echo state network (ESN) model, a form of recurrent neural network (RNN) with low computational complexity yet high efficiency, for turbulence channel prediction in FSO systems. We verify the prediction model using measured FSO channel data. Simulation results indicate that the ESN model has excellent prediction performance in FSO systems. Furthermore, we also highlight the effectiveness of the ESN model by comparing it with conventional CSI prediction models in terms of complexity and accuracy performance.