An End-to-End Autoencoder For FSO System Under Unknown CSI Scenarios

In free-space optical (FSO) communication systems, the maximum likelihood (ML) method is considered as the optimal detector when perfect channel state information (CSI) is available at the receiver. However, in scenarios where CSI is unknown, ML fails to provide satisfactory bit error rate (BER) performance. To address this issue, we propose an end-to-end (E2E) learning method for FSO communication based on convolutional autoencoder (CNN-AE). The proposed method is compared with a conventional FSO system using ML estimation in scenarios with perfect and unknown CSI, as well as varying levels of channel turbulence. Simulation results demonstrate that the proposed method outperforms the conventional ML method when CSI is unknown, while achieving equivalent performance to the ML detector under perfect CSI.