PINN-TurbNet for predicting optical turbulence in LEO satellite-to-ground laser communication links

Atmospheric optical turbulence (OT) remains a major limitation for the reliability of low Earth orbit (LEO) satellite-to-ground laser communication (SGLC) links. Here, we propose PINN-TurbNet, a physics-informed neural network that embeds planetary boundary layer (PBL) dynamics and Navier-Stokes constraints into a 3D Swin-Transformer U-Net. By jointly minimizing data and physics residuals, the model achieves a mean absolute error (MAE) of 0.313, outperforming conventional baselines. This framework provides a physically consistent and accurate approach for optimizing LEO-SGLC performance.