Deep Learning-Based Channel Modeling for Free Space Optical Communications

• DOI: 10.1109/jlt.2022.3213519
• Publication Date: 2022-10-11T19:31:15Z
• Authors: Wenbin Chen; Min Zhang; Dongdong Wang; Yueying Zhan; Shanyong Cai; Hui Yang; Zhiguo Zhang; Xue Chen; Danshi Wang
• Keywords: Free-Space Optical Communication; Modulation (music)

Instead of conventional knowledge-driven modeling, this study introduced data-driven channel modeling methods based on deep learning (DL) in a free-space optical (FSO) communication system to achieve low complexity, less difficulty, high fidelity, and differentiability. The FSO was modeled via three DL algorithms: generative adversarial network (GAN), bidirectional long short-term memory (BiLSTM), Bayesian neural (BNN). Consequently, their performance comprehensively compared terms amplitude waveforms, Kullback-Leibler (KL) divergence, 95% confidence interval, data distribution, fitted probability density functions. results indicated that the GAN-based exhibited optimal performance, implying its suitability for stochastic distribution channel. In addition, channels examined under different conditions: turbulence intensities (weak, moderate, strong), transmission distances (3, 4, 5 km), launch powers (0, 10, 20 dBm), modulation formats (on-off keying, pulse position modulation, 4), adaptability variation. test were satisfactory confirmed ability GAN learn approximate function Thus, primary aim prove with statistical characteristics, model can not only serve as an effective surrogate specific scenarios but also provide potential tool future methods.