基于观测器的机械臂位置/力神经网络控制  

Observer-based network control of manipulator position/force

作　　者：周芳[1] 朱齐丹[1] 蔡成涛[1] 赵国良[1] 

机构地区：[1]哈尔滨工程大学自动化学院,黑龙江哈尔滨150001

出　　处：《华中科技大学学报（自然科学版）》2009年第7期79-82,共4页Journal of Huazhong University of Science and Technology(Natural Science Edition)

基　　金：人事部出国留学人员基金资助项目(20001024A2)

摘　　要：针对机械臂末端运动受约束位置/力混合控制问题,提出一种基于观测器的神经网络自适应控制算法.假定机械臂非线性动力学模型未知,并且仅有机械臂关节角位置和接触力可测量,通过非线性坐标变换对位置和力控制进行解耦,得到降阶的位置和力动力学模型.利用高通滤波器对关节角速度进行重构,用神经网络对降阶未知动力学模型进行逼近,同时添加滑模控制项用于补偿神经网络的逼近误差,以改善系统的跟踪性能.最后,基于李亚普诺夫稳定性理论给出系统的稳定条件,并通过数字仿真验证了该算法的有效性.A neural network-based adaptive controller was proposed for the hybrid position/force control of the constrained robotic manipulator. The nonlinear dynamics of manipulator was assumed to be unknown and only the joint angle position and constrained force could be measured. The position/ force control could be decoupled by the nonlinear coordinate transformation, then a reduced position/ force dynamical model would be obtained. Among the algorithm, the high-pass filtering was utilized to estimate the manipulator joint angle velocity, while the neural network was employed to estimate the reduced unknown dynamic model, and the sliding mode item was used to compensate the estimated error in order to improve the tracking performance. At last, the stable conditions were obtained based on the theory of Lyapunov stabilization. The effectiveness of this control scheme was demonstrated by the simulation studies.

关 键 词：机械臂 位置/力混合控制 观测器 神经网络 滑模控制 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]