基于神经网络的人工肌非线性控制

运用非线性控制的逆系统方法，提出了一种基于神经网络的人工肌非线性控制方案，由原系统导出n-m阶逆系统模型，并与原系统一起构成具有反馈结构的伪线性系统，从而可方便地运用线性控制理论完成对控制系统的设计，用BP神经网络逼近逆系统模型，并借助于递推预报误差算法来训练神经网络，该算法与传统的BP算法相比具有更好的收敛特性，设计了一个具有单关节的人工肌试验系统，给出了人工肌关节跟踪正弦波和矩形波参考信号的试验结果，与传统的线性控制方案比较，基于神经网络的人工肌非线性控制方案能够得到更快的控制速度和更高的控制精度。

Nonlinear Control of Artificial Muscle Based on Neural Networks

Inverse system method for nonlinear control based on neural networks was employed to solve the nonlinearity of the pneumatic artificial muscles. The ( n-m )th order inverse system model which is deduced from the original system is linked with the original system to form a pseudo linear system. This system can be synthesized easily using linear control theory. Back propagation neural networks were introduced to approximate the inverse system model. A recursive prediction error algorithm which is much more effective than the back propagation algorithm was applied to train the neural networks. A single joint artificial muscle test system was designed. The experimental results show that the nonlinear control scheme is more effective than the traditional linear control scheme.