Analysis of deep learning models for rain-induced channel prediction in free-space optical communication system

Free-space optical (FSO) communication cites its advantages, such as high data rates, quick installation, economic affordability, and improved security. However, its drawback, mainly sensitivity to various weather phenomena, such as distortions and reflections by heat waves, dust, moisture, rainfall, fog, and scattering, degrades the performance of an FSO communication system. Goals 9 and 11 focus on industries, innovation, infrastructure, and sustainable cities to meet SDG 2030 in India. At the beginning of this paper, we propose an FSO communication system with different modulation schemes, from which we obtain the output power for different ranges as datasets. These are then trained using three different deep learning models, namely, gated recurrent unit, long short-term memory (LSTM), and recurrent neural networks, which are discussed in the later sections of the paper. The obtained results explain that the proposed system can achieve maximum accuracy and minimum absolute percentage error using the LSTM model. Furthermore, the prediction model LSTM has been carried out in various modulation schemes—on-off keying, quadrature phase shift keying, and quadrature amplitude keying—at data rates of 10, 20, and 125 Gbps to determine the accuracy of the model for the prediction of received signal strength indicator value under diverse rain attenuation.