An improved SPGD algorithm using a composite position-energy metric for incoherent beam combination

The conventional Stochastic Parallel Gradient Descent (SPGD) algorithm often fails in incoherent beam combination systems due to a vanishing gradient problem when initial pointing errors are large. To solve this, we propose an improved SPGD algorithm using a position-energy composite metric. This metric fuses a global centroid-based position signal with the standard Power-in-the-Bucket (PIB) signal. This dual-mode strategy employs the position metric for initial, wide-range ”capture” of off-target beams and then switches to the PIB metric for fine optimization of on-target power density. Numerical simulations under both no-turbulence and moderate-turbulence conditions validate that our algorithm achieves stable and rapid convergence from extremely large misalignments. Compared to the conventional method, it demonstrates a significantly enhanced capture range and robustness, while also effectively suppressing atmospheric turbulence-induced beam drift and spread. This research provides an efficient and reliable control solution for high-energy laser systems.