| 摘 要: 以模型为基础的传统的控制理论在设计控制器时需要预先知道被控对象的数学模型，而机械臂具有非线性、不确定的动态特性，因此应用传统的控制理论难以设计出适当的控制器。为解决此问题，本文应用自组织模糊径向基神经网络算法对机械臂的控制器进行设计，将径向基神经网络应用于自组织模糊控制算法，以及时调整学习率和权重分配，改善系统控制性能；应用L-M算法的概念，将所定义的误差函数最小化，搜索出神经网络参数值的修正量，以改进梯度下降法的收敛性，提升系统的控制品质。仿真结果表明，应用此算法，机械臂系统各轴的最大误差、均方根误差皆明显减小。 |
| 关键词: 机械臂；自组织模糊控制；径向基神经网络；自组织模糊径向基神经网络 |
|
中图分类号: TP241.2
文献标识码: A
|
| 基金项目: 天津市大学生创新训练计划项目(202010069122). |
|
| Design of Self-organizing Fuzzy Radial Basis Function Neural Network Controller for Manipulator System |
|
CHEN Tianyan, CHEN Qi, ZHANG Jing, HE Jingjing
|
(School of Information Engineering, Tianjin University of Commerce, Tianjin 300134, China)
2579028279@qq.com; chenqi@tjcu.edu.cn; 2889015652@qq.com; 1505293186@qq.com
|
| Abstract: Traditional model-based control theory needs to know the mathematical model of the controlled object in advance when designing the controller. However, the manipulator's nonlinear and uncertain dynamic characteristics make it difficult to design an appropriate controller using traditional control theory. To solve this problem, this paper proposes that self-organizing fuzzy radial basis function neural network algorithm is used to design controller of the manipulator. Radial basis function neural network is applied to self-organizing fuzzy control algorithm, so to adjust learning rate and weight distribution in time and improve the system control performance. L-M (Levenberg-Marquardt) algorithm is applied in the designing to minimize the defined error function. The correction amount of the neural network parameter value is searched to improve the convergence of the gradient descent method and improve control quality of the system. The simulation results show that by applying this algorithm, the maximum error and root mean square error of each axis of the manipulator system are significantly reduced. |
| Keywords: manipulator; self-organizing fuzzy control; radial basis function network (RBFN); self-organizing fuzzy radial basis function neural network (RBFNN) |
