摘要
提出了一种采用免疫粒子群优化算法对动态递归神经网络进行训练的方法,实现了对Elman网络的结构、权重、结构单元的初始输入和自反馈增益因子等参数的同时进化训练。进而针对非线性系统分别提出了相应的辨识与控制算法,并设计出了相应的辨识器和控制器。最后以超声马达为对象进行了仿真,结果表明:基于所提出的算法而设计的辨识器和控制器在辨识和控制过程中不仅都能取得很高的收敛精度和速度,而且对于随机扰动有较强的鲁棒性,从而为非线性系统的辨识和控制提供了一条新的途径。
A learning algorithm for dynamic recurrent Elman neural networks was presented, which is based on an immune particle swarm optimization （PSO）. The algorithm computed concurrently the evolution of network structure, weight, initial inputs of the context units and the self-feedback coefficient of the modified Elman network. Thereafter, a novel control method based on the proposed algorithm was introduced and discussed. More specifically, a dynamic identifier was constructed to perform speed identification, and a controller was designed to perform speed control for Ultrasonic Motors （USM）. Numerical experiments show that the identifier and the controller based on the proposed algorithm can both achieve higher convergence precision and convergence rate than those based on other state-of-the-art algorithms. In particular, the experiments show that the identifier can approximate the USM＇s nonlinear input-output mapping accurately. The effectiveness of the controller is verified using constant speed, step and sinusoidal changing speeds.
出处
《吉林大学学报（工学版）》
EI
CAS
CSCD
北大核心
2008年第4期858-864,共7页
Journal of Jilin University:Engineering and Technology Edition
基金
国家杰出青年科学基金项目(60625302)
“973”国家重点基础研究发展规划项目(2002CB3122000)
“863”国家高技术研究发展计划项目(2006AA04Z168)
国家自然科学基金项目(60433020)
关键词
人工智能
控制理论
动态递归神经网络
粒子群优化
免疫系统
超声马达
artificial intelligence
control theory
dynamic recurrent neural network
particle swarm optimization
immune system
ultrasonic motor
