机械手的在线鲁棒自适应神经网络跟踪控制  

On-line adaptive robust neural network tracking control for robot manipulators

作　　者：胡慧[1] 刘国荣[2] 

机构地区：[1]湖南大学电气与信息工程学院,湖南长沙410082 [2]湖南工程学院电气与信息工程系,湖南湘潭411104

出　　处：《控制理论与应用》2009年第3期337-341,共5页Control Theory & Applications

基　　金：湖南省自然科学基金资助项目(05JJ40093);湖南省科技计划项目(2008FJ3029).

摘　　要：考虑了一类具有外界干扰和不确定性的机械手臂轨迹跟踪鲁棒控制问题.控制器由自适应RBF(radial basis function)神经网络控制器和PD控制器组成.采用基于神经元灵敏度和获胜神经元概念的GP-RBF算法,在线确定神经网络的初始结构和参数.当误差满足一定要求时,根据Lyapunov稳定性理论的自适应律进一步调整网络权值,以保证机械手位置误差和速度跟踪误差渐近收敛于零.所设计的控制器可保证闭环系统的稳定性和鲁棒性.仿真结果证明了本文方法的有效性.The robust tracking control for a class of robot manipulators with disturbance and uncertainties is considered. The controller consists of an adaptive radial basis function（RBF） neural network controller and a PD controller. The initial structure and parameters of RBF neural network are determined on-line by the growing-and-pruning（GP-RBF） algorithm based on the sensitivity of neurons as well as the winner neuron concept. When the errors meet certain requirements, the adaptive law based on the Lyapunov stability further adjusts the weights of networks to ensure the asymptotic convergence of the tracking error to be zero. The controller guarantees the stability and robustness of the system. Simulation results demonstrate the efficacy of this method.

关 键 词：机械手臂 神经元灵敏度 获胜神经元 GP—RBF算法 轨迹跟踪 

分 类 号：TP241[自动化与计算机技术—检测技术与自动化装置]