一类不确定非线性系统反步自适应神经网络控制研究

Adaptive backstepping neural control for a class of uncertain nonlinear systems

针对含有高阶不确定扰动项且不可参数线性化的一类非线性系统 ,采用反步递推方法设计基于多层神经网络的自适应控制器 .多层神经网络可较好地逼近非线性系统 ,其权值能在系统先验知识不多的情况下在线调整 .给出了神经网络 L yapunov意义下稳定的在线自适应律 .在设计控制器的过程中 ,采用类加权形式 L yapunov函数 ,使得控制器能有效处理自适应控制奇异性问题 .仿真结果表明 ,该控制器对系统参数的不确定性和有界干扰具有一定的鲁棒性 。

A class of unknown nonlinear systems, which are not in parameter-linearizable expression with uncertain high-order disturbance, are considered. Based on backstepping approach, a multilayer neural network adaptive controller is presented for the nonlinear systems. Approximating nonlinear dynamic is one of the performances of multiplayer neural networks, and the NN weights are turned on-line without more prior knowledge of systems. The NN weight turn law is designed by Lyapunov synthesis approach, and the stabilization of the law is proved. Moreover, a novel quasi-weighted Lyapunov function is modified, which disposels effectively the issue of the singularity-free adaptive control. The simulation result shows that the controller is robust to some nonlinear uncertainties and bounded disturbance, and it can guarantee the global bound of all closed-loop signals.