基于RENN的柔性关节机械臂自适应动态面控制  

Flexible-joint manipulator adaptive control based on recurrent Elman neural networks and dynamic surface approach

作　　者：顾义坤 倪风雷 刘宏 

机构地区：[1]哈冬滨工业大学机器人技术与系统国家重点实验室,哈尔滨150001

出　　处：《控制与决策》2011年第12期1783-1790,共8页Control and Decision

基　　金：国家863计划项目(2008AA12A213);国家重点实验室自主部署课题项目(SKLRS200801A01)

摘　　要：针对面贴式永磁同步电机驱动的柔性关节机械臂动力学模型具有非线性、不确定性和未知外部扰动等特点,提出一种自适应动态面控制方法来实现其关节轨迹跟踪控制.控制律由动态面技术得到,降低了反推控制器的复杂性.模型不确定因素由递归Elman神经网络在线补偿,神经网络权值自适应律通过Lyapunov稳定性分析推导得到.仿真研究表明,该方法对于载荷不确定和外界扰动具有较强的鲁棒性,与传统动态面法相比,大大提高了柔性关节的位置跟踪精度.For overcoming the nonlinearity, uncertainty and unknown external disturbance in the model of flexible-joint manipulator driven by surface-mounted permanent magnet synchronous motors（PMSM）, an adaptive dynamic surface control（DSC） approach is proposed to design a position-tracking control system in the joint space. Control laws are obtained from DSC technique, which reduces the complexity of backstepping controller. The uncertainties of the model are observed and compensated online by the recurrent Elman neural networks（RENNs）. And the adaptation laws of RENNs＇ weights are induced from the Lyapunov stability analysis. The simulation studies show that the proposed method provides a good robustness against payload uncertainties and external disturbances, and the position tracking performance is improved greatly comparing with the conventional DSC method.

关 键 词：柔性关节机械臂 动态面控制 递归Elman神经网络 自适应控制 

分 类 号：TP24[自动化与计算机技术—检测技术与自动化装置]