摘要
针对多连杆柔性关节机械臂,设计了基于高斯径向基函数神经网络(GRBFNN)的滑模控制器,该控制器利用神经网络的逼近能力,将各关节的切换函数作为网络的输入,控制器完全由连续的RBF神经网络实现。利用该控制器与线性二次型跟踪器以及传统滑模控制器对三连杆柔性关节机械臂进行轨迹跟踪控制仿真。仿真结果表明:线性二次型跟踪器具有一定传输时延,滑模控制器跟踪轨迹有明显抖振,而神经滑模控制器取消了切换项,消减了抖振,具有良好的跟踪效果和稳定性。
A sliding mode controller for multiple-link flexible joint manipulator was proposed based on Gaussian Radial Basis Function Neural Networks （GRBFNN）. The switching function of each joint was regarded as the input of RBFNN, and the proposed controller was realized by continuous RBF neural network using its approximation ability. The simulation of tracking control of three-link flexible joint manipulator was done with the proposed controller, the linear quadratic regulator （LQR） tracker and traditional sliding mode controller （SMC）. The results show that the method of LQR tracker has some time delay, and the tracking trajectory has obvious chattering with the method of SMC. Due to cancel the switching item and eliminate chattering, neural sliding mode controllers have stability and keep the tracking effect well.
出处
《系统仿真学报》
CAS
CSCD
北大核心
2011年第10期2098-2102,共5页
Journal of System Simulation
基金
留学归国人员教学
科研建设项目
国家高技术研究发展计划(863计划)项目(2006AA04Z243)
