论文部分内容阅读
针对任意复杂的具有最小相位,滞后环节和非最小相位特性的离散非线性系统,提出一种通用的直接神经网络模型参考自适应控制。并采用具有在线学习功能的最近邻聚类算法训练RBF神经网络控制器,同时引入优化策略对聚类半径进行自动调整,并利用构造伪系统的方法构成一种对非最小相位同样有效的神经网络模型参考自适应控制器。仿真研究证明,该控制策略不仅能使多种非线性对象跟踪多种参考信号,而且抗干扰能力和鲁棒性也很好。
For any complex discrete nonlinear systems with minimum phase, hysteresis and non-minimum phase characteristics, a universal direct neural network model reference adaptive control is proposed. The nearest neighbor clustering algorithm with on-line learning is used to train the RBF neural network controller. At the same time, the optimization strategy is introduced to automatically adjust the clustering radius. The pseudo-system method is used to construct an equally effective neural network Model reference adaptive controller. Simulation results show that the proposed control strategy not only enables a variety of nonlinear objects to track multiple reference signals, but also has good anti-interference ability and robustness.