Properties of Multiobjective Robust Controller Using Difference Signals and Multiple Competitive Associative Nets in Control of Linear Systems
Properties of Multiobjective Robust Controller Using Difference Signals and Multiple Competitive Associative Nets in Control of Linear Systems
复制标题
DOI：
10.1007/978-3-319-12643-2_8
复制
发表时间：
2014-11
期刊：
影响因子：
--
通讯作者：
Weicheng Huang;Yuki Ishiguma;S. Kurogi
中科院分区：
文献类型：
--
作者：
Weicheng Huang;Yuki Ishiguma;S. Kurogi
最近，我们开发了多目标鲁棒控制器，使用非线性对象的差分信号，用于多个CAN 2学习和逼近非线性动力学的雅可比矩阵。在这里，CAN 2是一个人工神经网络，用于学习非线性函数的有效分段线性逼近。到目前为止，通过数值实验，我们已经表明，该控制器是能够应付的对象参数值的变化，以及通过切换多个CAN 2的控制目标的变化。然而，控制器还没有得到足够的分析。本文通过考察线性对象的控制，阐明了这种控制器的若干性质。
Recently, we have developed multiobjective robust controller using difference signals of nonlinear plant for multiple CAN2s to learn and approximate Jacobian matrices of the nonlinear dynamics. Here, the CAN2 is an artificial neural net for learning efficient piecewise linear approximation of nonlinear function. So far, by means of numerical experiments, we have shown that the controller is capable of coping with the change of plant parameter values as well as the change of control objective by means of switching multiple CAN2s. However, the controller have not been analyzed enough. This paper clarifies several properties of the controller by means of examining the control of linear plants.