基于随机并行梯度下降(SPGD)方法的自适应光学(AO)系统通过直接优化系统的性能评价函数来控制波前校正器以补偿光束中存在的波前畸变。为了提高这种无模型优化自适应光学系统的收敛速度, 提出了基于分区域耦合的新方法以改进传统随机并行梯度下降自适应光学系统的工作方式。将波前校正器光学孔径分成多块子区域, 每块子区域对应着的所有驱动器作为一个整体控制单元, 从形式上可以得到一个空间分辨率较低的分区域波前校正器。该校正器与原校正器同步工作, 并采用随机并行梯度下降算法对同一个性能评价函数进行优化, 从而构成了双校正器的耦合工作结构。对256单元分立活塞式波前校正器建立了自适应成像系统的数值模型, 结果表明这种分区域耦合的随机并行梯度下降自适应光学系统比传统随机并行梯度下降自适应光学系统具有更快的收敛速度和更好的渐近态。

Adaptive optics (AO) system based on stochastic parallel gradient descent (SPGD) method compensates a beam′s wave-front aberrations by controlling a wave-front corrector with direct optimization of the system′s performance metric. A new method of subzones coupling for modifying the conventional SPGD AO system is presented to increase the convergence speed of the model-free adaptive optics system. A new lower-resolution subzonal wave-front corrector can be approximated by subdividing the wave-front corrector′s aperture into multiple subzones, all actuators associated with each subzone being treated as a whole controlling element. The coupling architecture of double wave-front correctors can be constructed by simultaneously controlling the approximated wave-front corrector and the original one with SPGD technique for a common performance metric. A numerical simulation model for an adaptive imaging system with 256-elements pixelated piston-type wave-front corrector was built up. The numerical simulation results show that this SPGD AO system based on subzones coupling holds faster convergence speed and better asymptotic normality than the conventional SPGD AO system.