非线性系统神经网络控制器 BP 算法的改进

Improved BP Algorithm of Neural Network Controller of Nonlinear System

对非线性系统神经网络控制器提出了一种改进的误差反向传播算法。该算法引入一维参数寻优法，预估神经网络连接权系数学习步长，减少了步长选取的盲目性，避免了学习过程陷入局部极小的可能性，并以二自由度机器人为对象，进行了神经网络控制的仿真研究。

An improved Backward Error Propagation Algorithm used as neural network controllers to control a nonlinear system is presented in this paper.In order to make the coefficient converges at overall area(not at local minimum value),one dimension parameter optimization method to estimate the step length of the coefficient is adopted. The approach is applied to a two free degree robot manipulator and good simulation results are obtained.