论文部分内容阅读
在选择节点激活的基础上,设计利用动态RBF逼近非线性函数的神经网络自适应控制器,提出了催眠/唤醒(激活)技术,由于对动态网络中的某些节点进行了催眠,从而使网络规模最小.依此设计仿射非线性系统自适应控制器,得到了基于LYAPUNOV稳定性定理的网络权值更新律,同时引入了滑模控制,对由网络内在逼近误差、忽略不活动节点及暂时睡眠节点所引起的扰动项进行补偿,这样就确保了系统的稳定性.仿真结果表明,与仅采用选择节点激活技术的控制器相比,采用催眠/唤醒(激活)技术的控制器具有更优的跟踪性能、更小的网络规模和更短的执行时间.“,”The learning rate of an adaptive controller based on RBF (Radial Base Function) network is somewhat higher than that of an adaptive controller based on BP network. S.Fabri proposed a significantly smaller dynamic RBF network adaptive controller. We believe that Fabri′s proposal can be further improved. We propose a new hypnotizing/waking(activating) technique, which is used to design a dynamic RBF network adaptive controller. The minimum size of network is reached as the technique selects and hypnot...