Deep Learning Based Free Space Optical Communication Diversity System

In this paper, a DL-based spatial diversity system for free-space optical (FSO) communication is proposed. Compared to traditional FSO spatial diversity systems that employ Maximum Ratio Combining (MRC), Equal Gain Combining (EGC), and Selection Combining (SC) techniques, the proposed system reduces system complexity while maintaining an acceptable performance. When simulating atmospheric turbulence with 16 turbulence states, the DL-based FSO diversity system achieves an operating time of 5.56s, whereas the traditional FSO diversity system requires 8.24s under the same conditions. In fact, turbulence states in the range of 200–300 are commonly employed to better emulate atmospheric channels. As the number of turbulence states increases, the DL-based FSO diversity system demonstrates a more significant performance advantage in terms of operating time compared to traditional systems.