自适应光学系统变形镜控制电压预测

在校正大气湍流畸变波前相差的自适应光学系统中，利用基于Levenberg-Marquardt学习算法的非线性反向传播神经网络技术（LMBP）对变形镜控制电压进行预测。以对受横向风影响的大气湍流畸变波前斜率数据为研究对象，通过数值仿真方法，研究了基于LMBP算法的自适应光学系统变形镜电压非线性预测控制算法。通过实验发现，预测电压和变形镜实际控制电压拟合效果良好。讨论了回溯帧数对预测效果的影响，并与基于递推最小二乘（RLS）算法的线性预测算法进行比较。对比结果表明，基于LMBP算法的非线性电压预测方法比基于递推最小二乘法的线性电压预测方法能更有效地降低系统由伺服延迟引起的误差。

Predicting control voltages of deformable mirror in adaptive optical system

The paper proposes a nonlinear neural network algorithm to predict the control voltages of deformable mirror in adaptive optical systems. The algorithm, named LMBP algorithm, adopts the two layer back propagation neural network based on Levenberg Marquardt learning algorithm. The numerical simulation with the algorithm has been conducted by using the gradient data of atmosphere turbulence wavefront influenced by transversal wind. The predicted control voltage agrees well with the actual control voltage of deformable mirror. The influence of the look back frame number on the prediction effect is discussed and the residual error of the adaptive optic system is calculated with the LMBP nonlinear prediction algorithm and the recursive least square (RLS) algorithm. The residual error caused by servo lag in the system is reduced more effectively using the LMBP nonlinear prediction algorithm.