论文部分内容阅读
自适应光学系统的性能受限于伺服系统的延迟误差和波前传感器的光电子噪声。提出了一种多模型单变量预测模型,该模型采用基于Levenberg-Marquardt学习算法的前馈型神经网络。利用计算机多核处理器,设计了一个具有并行处理能力的预测控制器,来实现对自适应光学闭环控制电压的预测,以消除延迟误差的影响。通过数值仿真实验,研究了预测控制器对控制电压和远场斯特雷尔比的影响,与未采用预测控制器的系统进行了比较,并对预测算法的并行性能进行了分析。实验结果表明,使用并行化方法的预测控制器可以有效缩短系统的预测时间,提高预测算法的加速比,与经典比例积分(PI)控制算法相比可以更有效地降低系统由于伺服延迟引起的误差,远场的斯特雷尔比有明显地提高。
The performance of the adaptive optics system is limited by the servo system’s delay error and the photoelectron noise of the wavefront sensor. A multi-model univariate prediction model is proposed, which uses a feedforward neural network based on Levenberg-Marquardt learning algorithm. Using computer multi-core processor, a predictive controller with parallel processing capability is designed to predict the adaptive optical closed-loop control voltage to eliminate the influence of delay error. The influence of the predictive controller on the control voltage and the far-field Stratal ratio is studied by numerical simulation. Compared with the system without the predictive controller, the parallel performance of the predictive algorithm is analyzed. The experimental results show that the predictive controller using the parallelization method can effectively reduce the prediction time of the system and improve the speedup of the prediction algorithm, which can reduce the system error due to servo delay more effectively than the classical proportional integral (PI) control algorithm , Farrell than the far field has significantly improved.