为提高基于随机并行梯度下降(SPGD)算法的高分辨率自适应光学(AO)系统的有效校正带宽, 提出了分级SPGD波前校正的方法。在每一级SPGD波前校正中, 将高分辨率波前校正器的控制单元按位置分成多组, 每一组都施加相同的控制电压作为一个控制变量。采用的校正级越高, 独立的控制变量的数目也越多。多级SPGD自适应光学系统按从低到高的等级顺序对畸变波前进行校正。以一个具有16×16控制单元的分立活塞式波前校正器为基础建立了基于3级SPGD波前校正的自适应光学系统的数值模型, 并针对大气湍流引起的某一帧随机相位屏做了模拟校正实验。结果表明采用分级波前校正后收敛速度提高了23%, 此方法确实提高了基于随机并行梯度下降算法的自适应光学系统的有效校正带宽。

A multilevel wavefront correction technique is proposed to improve the effective correction bandwidth of a high-resolution adaptive optics system based on stochastic parallel gradient descent (SPGD) algorithm. At every level of wavefront correction based on SPGD, the control elements of the high-resolution wavefront corrector are subdivided into many clusters according to their locations. The control elements in each cluster are driven by the same control voltage which is treated as a control variable in SPGD algorithm. The higher the correction level is, the bigger the number of independent control variables is. The multilevel SPGD adaptive optics system corrects the distorted wavefront in the sequence from low level to high level. The numerical model of a 3-level SPGD adaptive optics system with a pixelated piston-type wavefront corrector having 16×16 control elements was built, and the simulation correction experiments for a certain stochastic phase screen introduced by atmospheric turbulence were performed on the system. The numerical simulation results demonstrate that the convergence rate of multilevel SPGD wavefront correction is increased by 23% than that of conventional SPGD wavefront correction. It indicates that the multilevel wavefront correction technique can indeed improve the effective correction bandwidth of adaptive optics system based on SPGD algorithm.