基于多层神经网络的饱和非线性输入自适应控制  

Adaptive Control with Saturation Nonlinear Input Based on Multi-Neural Networks

作　　者：李红春[1] 梅建东[2] 任云晖[1] 

机构地区：[1]江海职业技术学院,扬州225101 [2]扬州职业大学,扬州225009

出　　处：《扬州职业大学学报》2011年第1期33-36,39,共5页Journal of Yangzhou Polytechnic College

摘　　要：针对一类具有未知饱和模型的单输入单输出非线性系统的控制问题,根据滑模控制原理和多层神经网络的逼近能力,提出了一种直接自适应神经网络控制器的设计新方案。该方法将控制增益推广到未知函数,通过补偿饱和模型方法取消了饱和模型各参数已知的条件。鲁棒项的引入消除了建模误差和参数估计误差的影响。理论分析证明了闭环系统是半全局一致终结有界,跟踪误差收敛到零的邻域内。仿真结果进一步表明所提控制方法的有效性。The nonlinear system with saturation input is discussed in this paper.Based on the principle of sliding mode control and the approximation capability of multi-neural networks,a new design scheme of adaptive neural network controller is proposed.The control gain is extended to unknown function and the condition of known parameters is canceled by compensating saturation.Robust term is used to eliminate the approximating error and disturbance.By Lyapunov function,the closed-loop system is shown to be uniformly ultimately bounded and tracking error asymptotically converges to zero.Simulation results show the effectiveness of this approach.

关 键 词：自适应控制 神经网络 饱和 非线性控制 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]