基于小波神经网络的机器人模糊控制算法

Robot fuzzy control algorithm based on wavelet neural network

为了提升工业制造过程中工业机器人机械臂的控制精度,文中基于变换域的思想设计了一套智能控制算法。该方法采用小波变换函数替代了传统神经网络隐藏层中的非线性激活函数,以获得紧密型的小波神经网络,从而提升了网络对于弱信号的特征提取能力。同时,将小波网络提取的特征输入至模糊控制网络中,并引入T-S推理规则,保证了算法对于复杂过程的辨识及控制能力。同时算法基于误差的反向传播理论进行训练,可以灵活调整学习率和迭代速度,确保了该过程的收敛性。以二阶倒立摆系统为控制对象进行仿真实验,根据不同变量对系统影响力的不同设计了两级控制网络,以增强网络的泛化能力。仿真结果表明,所提算法控制下的二阶倒立摆稳定性明显优于BP神经网络,其位移与上、下摆角等关键参数的方差分别提升了54.72%、56.32%和51.04%。

In order to improve the control accuracy of robots in industrial manufacturing process,a set of intelligent control algorithm is designed based on the idea of transformation domain.In this method,the wavelet transform function is used to replace the nonlinear activation function in the hidden layer of the traditional neural network,and a compact wavelet neural network is obtained,which improves the feature extraction ability of the network for weak signals.At the same time,the features extracted by wavelet network are input into the fuzzy control network,and T-S inference rules are introduced to ensure the identification and control ability of the algorithm for complex processes.The algorithm is trained based on the back propagation theory of error,which can flexibly adjust the learning rate and iteration speed of training,and ensure the convergence of the convergence process.The simulation experiment is carried out with the second-order inverted pendulum system as the control object.According to the different influence of different variables on the system,a two-level control network is designed,which improves the generalization ability of the network.The simulation results show that the stability of the second-order inverted pendulum under the control of the proposed algorithm is significantly better than that of the BP neural network,and the variance of its key parameters such as displacement,up swing angle and down swing angle are increased by 54.72%,56.32%and 51.04%respectively.