Multi-Link Turbulence Sensing and Mitigation for FSO Diversity Reception With Attention-Based Neural Combining

With the rapid development of applications such as sixth-generation (6G) communications, the low-altitude economy, and low-earth-orbit (LEO) satellite systems, the demand for high-capacity wireless communications is becoming increasingly urgent. Free-space optical (FSO) communication is a key enabling technology due to its large bandwidth potential, yet FSO links are highly vulnerable to atmospheric turbulence, which severely constrains system capacity and reliability. This paper proposes a multi-link FSO closed-loop sensing-assisted communication framework. Reverse-propagating probing beams are used for turbulence sensing, and a spatial light modulator (SLM)-based optical phased-array independently controls the beam on each link to suppress turbulence. At the receiver, a self-attention neural network is employed for signal combining to improve diversity reception performance. Two independent turbulence-impaired links with different propagation directions and lengths of 1.2 m and 1.3 m are implemented in the experiment. Nonuniform turbulence with intensity fluctuations up to 10 dB is emulated using SLMs, and the sensing error is below 0.5 dB. With multi-beam control, the average received optical power increases by 2.92 dB for the 1.2-m link and 2.85 dB for the 1.3-m link, and the scintillation index is reduced by 82.7% and 81.3%, respectively. In a four-wavelength wavelength-division-multiplexed (WDM) transmission with an aggregate rate of 256 Gb/s, the system robustness improves from a minimum of 31.7% for a single link to 100% after dual-link diversity combining. The results demonstrate that the proposed technique can significantly enhance the capacity and reliability of FSO communications and is promising for practical applications.