ESN网络在机器人轨迹跟踪控制中的应用  

Application of the Echo State Network(ESN) in Trajectory Tracking Control for Robots

作　　者：杜佩君[1] 

机构地区：[1]兰州交通大学自动化与电气工程学院,甘肃兰州730070

出　　处：《自动化仪表》2013年第6期4-8,共5页Process Automation Instrumentation

摘　　要：为了研究具有模型不确定性的机器人操作手的轨迹跟踪控制,采用一种新的递归神经网络——回声状态网络(ESN)设计了动态控制器。采用PID控制器补偿ESN网络的逆建模误差,并在网络训练过程中加入白噪声项,以保证动态系统的稳定性。最后针对两关节机械手的轨迹跟踪控制问题进行了数值仿真,仿真结果表明了该方法的有效性。To research the trajectory tracking control of robotic manipulator which features model uncertainty, by adopting the new type of recurrent neural network, i. e. , echo state network （ ESN ）, the dynamic controller is designed. The inverse modeling error of the ESN is compensated by PID controller, and during network training, white noise term is added for ensuring the stability of the system. Aiming at the trajectory tracking control of two-joint manipulator, numerical simulation is carried out. The simulation results show the effectiveness of this method.

关 键 词：神经网络 回声状态网络(ESN) 机器人 轨迹跟踪 自适应控制 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]