GPU-ACCELERATED TURBULENCE SIMULATOR FOR SPACE-BASED OPTICAL COMMUNICATIONS

Abstract

In modern space-based networks, growing bandwidth demands and increasing mission complexity drive the need for accurate, high-resolution simulations that capture small-scale turbulence features and rapid satellite movement. By leveraging GPU parallelization, this simulator manages large data volumes and frequent time steps, enabling the modeling of wavefront distortions essential for robust system design, adaptive optics, and performance optimization. This thesis presents a novel, highspeed simulator for optical propagation through dynamic atmospheric turbulence affecting satellite downlinks. The method begins by calculating the effective refractive index structure parameter (C 2 n ) which captures the turbulence strength, along the observation path between the satellite and ground receiver. This derived C 2 n informs the atmospheric slicer, which distributes phase screens throughout the propagation path. The simulation focuses on the first 20 km of atmosphere, where the majority of turbulence affecting free-space optical links occurs. The angular spectrum propagation formula is implemented to achieve Fresnel propagation between planes where phase screens represent integrated turbulence slices at specific altitudes. Temporal evolution is achieved via the frozen flow hypothesis and an adjustable wind model with altitude-dependent wind speeds. Numerical simulation of optical propagation through turbulence with high spatial sampling poses significant computational challenges, especially for rapidly moving Low Earth Orbit (LEO) satellites and simulations with numerous phase screen layers. This work addresses this challenge with an innovative GPU architecture that parallelizes intensive computations and large loops across GPU cores. This advancement enables the use of large phase screens, many layers, and rapid propagation loops, efficiently simulating fast-translating LEO satellites. This comprehensive approach significantly enhances the speed of atmospheric turbulence simulations for satellite communications, offering a powerful tool for system design, performance prediction, and optimization of adaptive optics strategies in free-space optical communication systems. The GPU-accelerated implementation achieves speedup factors of 310× to 600× over conventional CPU-based simulators.