Accelerating SPGD-AO Compensation: A High-Performance FPGA-Based Approach for Optical Communication

The Stochastic Parallel Gradient Descent (SPGD) algorithm combines the advantages of stochastic gradient descent and parallel computing, and has been widely applied in adaptive optics. However, traditional SPGD algorithms suffer from long operation cycles and numerous iterations, making it difficult to meet the real-time requirements of adaptive optical systems in certain application scenarios. To address this issue, this paper proposes an FPGA-based adaptive optics scheme optimized with the Adam algorithm. By integrating the Adam optimizer to enhance the convergence efficiency of SPGD while leveraging the high-speed parallel computing and low-power characteristics of FPGAs, the proposed design significantly improves the system's real-time processing capability while effectively correcting beam distortions caused by atmospheric turbulence. Simulation results demonstrate that, compared with the MATLAB-based adaptive optics approach, the proposed system reduces the runtime per iteration from 501 ms to 5.7 ms with an error of approximately 0.1%, thereby verifying its accuracy and efficiency. This work provides an effective solution for developing high-performance and low-latency adaptive optics systems.