基于回归神经网络的滑模跟踪器设计

基于回归神经网络的在线辨识 ,为实现非线性系统自适应控制提供了一种很有应有前景的方法。本文基于具有线性输入特性的神经网络 ,提出了使系统辨识误差具有指数收敛特性的学习算法。为了得到尽可能普遍的控制律 ,本文运用滑模技术设计出控制信号 ,用其补偿神经网络模型与非线性系统之间的匹配误差。以此为基础 ,提出一种新的基于回归神经网络的间接自适应控制方案。仿真结果表明 ,本控制方案具有较好的跟踪性能

Design of a Trajectory Tracking Controller Based on Sliding Mode Technique and Recurent Neural Network

The on line identification based on the recurrent neural network provides a promising method for realizing the adaptive control of a non linear system. This paper presents a learning algorithm which makes the identifying error converge to zero at an exponential rate. The method is based on neural networks which have linear inputs. In order to get a more generalized control law, the sliding mode technique is used to design control signals which can compensate for the matching errors between the neural networks and the non linear system. Accordingly, an indirect adaptive controller based on recurrent neural network is presented. The effectiveness of the scheme is testified by simulation.