全局自适应神经网络跟踪控制及逼近域确定

Globally adaptive neural network tracking control and determination of approximation domain

针对自适应神经网络跟踪控制问题,提出一种确定逼近域的方法.采用参考信号取代未知非线性函数中的系统输出,神经网络用于逼近以参考信号为输入的未知不确定项.可以利用参考信号的界预先确定神经网络逼近域,再采用自适应鲁棒方法处理由于函数输入置换所引起的另一类不确定项.所得到的闭环系统是全局稳定的.仿真实例说明了该控制方法的有效性.

A method to determine the neural network approximation domain is developed for adaptive neural network tracking control problem. The system outputs in unknown nonlinear functions are replaced by the reference signals so that neural networks are employed to approximate the unknown uncertainties whose inputs are the reference signals. The designer can determine neural network approximation domain based on the bound of the reference signals. The adaptive robust technique is used to handle the other kind of uncertainties which results from the replacements of function inputs. The closed-loop system is proved to be globally stable. A simulation example shows the effectiveness of the control method.