Language selection

Government of Canada / Gouvernement du Canada

Search

GPU-accelerated multilayer turbulence simulation for modeling of space-to-ground optical links

DOIResolve DOI: https://doi.org/10.1109/ICSOS66026.2025.11443182
AuthorSearch for: 1; Search for: 1; Search for: 2ORCID identifier: https://orcid.org/0000-0001-5937-6254; Search for: 1
Affiliation
  1. McMaster University
  2. National Research Council Canada. Digital Technologies
FunderSearch for: National Research Council Canada
FormatText, Article
Conference2025 IEEE International Conference on Space Optical Systems and Applications (ICSOS), October 28-31, 2025, Kyoto, Japan
Subjectfree-space optical communication; atmospheric turbulence; GPU acceleration; split-step Fourier propagation; phase screen; TensorFlow
Abstract
Satellite-to-ground free-space optical links are limited by atmospheric turbulence that distorts the received wave-front and induces scintillation. We present a GPU-accelerated multilayer turbulence simulator that models the dynamic, altitude-stratified propagation for space-to-ground link geometries. The simulator is based on standard Kolmogorov phase screen and split-step Fourier propagation, but implemented in TensorFlow for significant parallelization and accelerated computation time. The simulation stores propagation operators and covariance terms as GPU tensors to avoid transfer overhead and enable high throughput at significantly reduced computation time compared with freely available turbulence propagation tools. Using a single mid-range gaming GPU the simulator was able to achieve up to 310 times speed up over existing tools, with better scaling to larger grid sizes. This acceleration enables larger-aperture, long time series, and high resolution simulations of downlink geometries for machine-learning-driven adaptive optics and signal-processing mitigation in next-generation FSO systems.
Publication date
PublisherInstitute of Electrical and Electronics Engineers
In
  • 2025 IEEE International Conference on Space Optical Systems and Applications (ICSOS), 11443182 (24 March 2026): 1–6.
LanguageEnglish
Peer reviewedYes
Export citationExport as RIS
Report a correctionReport a correction (opens in a new tab)
Record identifier06dcbc23-ff04-4304-b3e8-006b347226fe
Record created2026-04-16
Record modified2026-06-03

Page details

From:

Date modified: