光学学报, 2014, 34 (s2): s214003, 网络出版: 2014-11-17
一种改进的全局耦合的随机并行梯度下降算法
Improved Method of Stochastic Parallel Gradient Descent Algorithm with Global Coupling
光计算 自适应光学 随机并行梯度下降算法 全局耦合 波前畸变补偿 optics in computing adaptive optics stochastic parallel gradient descent algorithm global coupling wavefront correcting
摘要
基于随机并行梯度下降(SPGD)算法的自适应光学(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波前畸变。传统的全局耦合的SPGD算法能够明显提高AO系统的收敛速度,但会降低系统的波前校正效果。针对传统算法的不足提出了改进方法:收敛前后期采用不同的电压优化控制方式。建立了SPGD AO成像系统的仿真模型,结果表明,与传统全局耦合SPGD算法相比,改进算法在收敛速度基本与前者相同的情况下,波前校正后焦平面成像光斑的斯特列尔比(SR)提高了0.14。
Abstract
The stochastic parallel gradient descent (SPGD) algorithm adaptive optics (AO) system controls deformable mirror to correct the distorted wavefront by optimizing quality metric function of the system directly. The global coupling SPGD algorithm can significantly increase the convergence rate of the system, while slightly declining the performance of wavefront correcting. Thus an improved method to solve the problem is proposed. To demonstrate this method, a simulated model of SPGD AO system is built. The simulated result indicates that the improved algorithm not only retains the convergence rate of conventional global coupling SPGD algorithm but also increases the Strehl rate (SR) of the corrected image on the focal plane by 0.14.
蒋鹏志, 梁永辉, 许洁平, 毛宏军. 一种改进的全局耦合的随机并行梯度下降算法[J]. 光学学报, 2014, 34(s2): s214003. Jiang Pengzhi, Liang Yonghui, Xu Jieping, Mao Hongjun. Improved Method of Stochastic Parallel Gradient Descent Algorithm with Global Coupling[J]. Acta Optica Sinica, 2014, 34(s2): s214003.