神经网络结合并行混沌粒子群优化算法的波前预测器  

A Novel Neural Network Wavefront Predictor Based on Parallel Chaotic PSO Algorithm

作　　者：白亚男[1] 史永昌[2] 史晓雨[3] 

机构地区：[1]平顶山学院软件学院,平顶山467000 [2]平顶山学院计算机科学与技术学院,平顶山467000 [3]电子科技大学计算机科学与工程学院,成都611731

出　　处：《计算机与数字工程》2013年第6期1024-1026,共3页Computer & Digital Engineering

基　　金：基于无线自组网的应急通信关键技术问题研究(编号:61072043)资助

摘　　要：由伺服系统所引起的延迟误差很大程度上影响着自适应光学系统(AOS)的性能。为了提高神经网络波前预测器的收敛性和精度,论文提出了一种改进的并行混沌粒子群优化算法(PCPSO)。该算法利用混沌序列的随机和遍历特性,以及并行化处理方法,有效地避免早熟现象的发生。通过对基准函数的测试,比起经典粒子群算法,该算法具有较高的计算速度和收敛能力。并且,论文设计了一种基于PCPSO算法的神经网络波前预测器,并运用于自适应光学系统的波前预测中。通过数值仿真实验,研究了这种新型波前预测器的性能,实验结果表明,采用并行化混沌粒子优化算法的预测器可以有效的预测控制电压信号,具有很好的泛化能力,对提高光束质量起到了重要的作用。A delay caused by the servo system error greatly affects the performance of the adaptive optics system（AOS）. In order to im- prove the convergence and accuracy of the wave-front of the neural network to predict, this paper presents an improved parallel chaos particle swarm optimization algorithm（PCPSO）. The algorithm uses the characteristics of the chaotic sequence of random and traverse, and the paral- lel processing method to avoid premature phenomenon occurred. The benchmark function test, compared to the classical particle swarm opti- mization, the algorithm has a high computational speed and convergence. And design a neural network wavefront based on PCPS（） algorithms predict, and applied to the adaptive optics system wave forecast. By numerical simulation, experimental results show that, to using parallel- ization chaos predicted particle optimization algorithm can effectively predict the control voltage signal, with good generalization ability, and has played an important role in improving the beam quality.

关 键 词：人工神经网络 并行 波前预测器 粒子群优化 混沌 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]