摘要
基于对象的逆动力学模型,提出一种RBF神经网络逆控制的在线自学习控制方案。辨识器采用RBF神经网络和最近邻聚类学习算法,实现对象逆模型的动态辨识。并引入优化策略对聚类半径进行自动调整,以保证聚类的合理性。为克服逆控制对非最小相位系统的不足,利用构造伪系统的方法,构成一种对非最小相位系统仍然有效的神经网络逆控制器。仿真实验证明该控制策略不仅能使非线性非最小相位系统具有良好的动态跟踪性能,而且具有较好的抗干扰能力。
Based on the inverse system model, a method of on-line self-learning control strategy was proposed, which used inverse control based on RBF neural network. The system identification based on RBF neural network which applied nearest neighbor clustering algorithm realized the identification of the inverse dynamic system model. The optimization strategy which regulated the clubman radius automatically was introduced to guarantee the rationality of clubman. Through constructing pseudo-plant, a neural network inverse controller is put forward which is still effective to the nonlinear non-minimum phase system. With the help of simulation, the control strategy can not only have a favorable dynamic tracking performance to nonlinear non-minimum phase system, but also resistance to disturbance of system.
出处
《系统仿真学报》
CAS
CSCD
北大核心
2009年第4期962-964,968,共4页
Journal of System Simulation
基金
国家自然科学基金面上项目(50775110)
关键词
非线性非最小相位系统
最近邻聚类算法
神经网络逆控制
在线自学习
nonlinear non-minimum phase system
nearest neighbor clustering algorithm
neural network inverse controller
on-line self-learning