自适应光学系统几种随机并行优化控制算法比较
Comparison of several stochastic parallel optimization control algorithms for adaptive optics system
-
摘要: 直接对系统性能指标进行优化是自适应光学系统中一种重要的波前畸变校正方法，选择合适的随机并行优化控制算法是该技术成功实现的关键。以32单元变形镜为校正器，基于多种随机并行优化算法建立自适应光学系统仿真模型。从算法的收敛速度、校正效果、局部极值3个方面对遗传算法、单向扰动随机并行梯度下降、双向扰动随机并行梯度下降及模拟退火算法进行了比较。仿真结果表明，遗传算法收敛速度太慢，不适用于需要实时控制的自适应光学系统；双向扰动随机并行梯度下降算法收敛速度、校正效果要优于单向扰动随机并行梯度下降，且能够适应各种情况下的扰动电压；模拟退火几乎以概率1收敛到全局极值附近，且收敛速度是上述算法中最快的。
-
关键词:
- 自适应光学系统 /
- 随机并行梯度下降算法 /
- 模拟退火 /
- 遗传算法 /
- 数值仿真
Abstract: Optimizing the system performance metric directly is an important method for correcting wave-front distortions in adaptive optics(AO) systems. Appropriate stochastic parallel optimization control algorithm is the key to correcting distorted wave front successfully. Based on several stochastic parallel optimization control algorithms, an adaptive optics system with a 32-element deformable mirror was simulated. Genetic algorithm(GA), the unilateral perturbation stochastic parallel gradient descent (SPGD), the bilateral perturbation SPGD and simulated annealing(SA) were compared in convergence speed, correction capability and local maximum. The results show that because of the unaceptable convergence speed, GA is not suitable for the control of real-time AO system; the bilateral perturbation -
点击查看大图
计量
- 文章访问数: 2227
- HTML全文浏览量: 204
- PDF下载量: 840
- 被引次数: 0
下载:
