基于新型动态神经元网络的逆系统方法

Inverse System Control Using New Dynamic Neural Networks

针对提高逆系统建模中神经网络的逼近效果和动态性能问题,根据PID神经元网络工作原理,提出一种具有动态激励函数的新型PID神经元模型—输出反馈型PID神经元(OFPID),输出激励采用连续的Sigmoidal函数,使神经元具有等效的IIR突触,采用梯度下降法实现OFPID神经元网络的权值调整,将其应用于非线性系统的神经网络逆控制系统,从而提高非线性系统的解耦效果和控制性能。仿真实验证明,提出的新型神经元网络是一种良好的非线性系统建模和控制工具。

In order to improve the approximation power and the response performance of neural networks in inverse dynamic control schemes, a new type artificial neuron, called OFPID artificial neuron, is proposed according to principle of PID artificial neuron. It works as a dynamic mapping, and has equivalent IIR synaptic and output excitation of Sigmoidal function. Weights of OFPID neural networks are adjusted by gradient descent method, and the neural networks are used to model the inverse dynamics of nonlinear systemes. Simulation results show that the OFPIDNN can achieve much better dynamic performance, and this implies that the proposed new neural network model has good potential in nonlinear modeling and control.