基于神经网络的非线性自适应输出反馈控制

针对能够采用仿射非线性表示的含有未建模动态的SISO非线性系统，讨论了一种基于神经网络的自适应控制方法，该方法对受控对象的已知部分，有用反馈线性化方法设计控制器，用神经网络在线补偿未建模动态及外部干扰等引起的误差，从而实现自适应控制。对具有未建模动态的双车倒立摆设计了输出反馈自适应控制系统，仿真表明该方法是有效的。

Nonlinearly Adaptive Output Feedback Control Using Neural Network

A discussion is devoted to design neural network adaptive control scheme of the SISO (single input and single output) nonlinear system with unmodeled dynamics. According to the known part of the plant, feedback linearization method is used to design the controller. The error resulted from the unmodeled dynamics and the external disturbance is compensated by online neural network. The neural networks are designed as a five layer fuzzy neural network and its construction is optimized by genetic algorithms. It has been used to approtimate the nonlinear function of system and to compesate the error of unmodeled dynamic. The design of neural network adaptive controller has better performances. The method is verified by the digital simulation of two cart with inverted pendulum system and unmodeled dynamics.