基于增广ESN的机器人轨迹跟踪控制  

Trajectory Tracking Control for Robots Based on Augmented ESN

作　　者：杜佩君[1] 张瑞锋[2] 

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 [2]哈尔滨工业大学航天学院,黑龙江哈尔滨150001

出　　处：《信息与控制》2013年第4期443-448,共6页Information and Control

摘　　要：研究了具有模型不确定性的机器人操作手轨迹跟踪控制问题.针对以往动态递归神经网络(recurrent neural network,RNN)训练算法难以实现及机械手的强非线性等问题,提出一种基于"增广"策略的回声状态网络(echo state network,ESN)方法 A-ESN(augmented echo state network),网络使用"增广"学习策略离线训练ESN输出权值,训练过程中加入服从均匀分布的白噪声项来保证动态系统稳定性.针对两关节机械手的轨迹跟踪控制问题,首先用A-ESN辨识机械手不确定部分的逆模型,并用PID反馈控制器补偿A-ESN网络的逆建模误差;然后基于A-ESN设计动态控制器;最后进行了数值仿真,并与常规递归神经网络算法进行了比较,结果显示该方法的控制精度比常规方法有了很大提高,表明了该方法的有效性.The trajectory tracking control of robot manipulator with model perturbation is studied. For the difficulties in gaining algorithms of the existing recurrent neural network （RNN） and high nonlinearity of the manipulators, an augmented echo state network （A-ESN） is proposed. The output weights of the ESN is trained off-line and the ＂augmented＂ strategy are used. A small uniform white noise term is added during the training in order to stabilize the dynamic system. For the trajectory tracking control of two-joint manipulator, the uncertainty of manipulators is firstly identified by an A-ESN, and a PID feedback controller is added to compensate the inverse modeling error. Then a dynamic controller is designed based on A-ESN. Finally, a numerical simulation is carried out. The simulation results show that compared with conventional recurrent neural network algorithms, the control precision of this method has greatly improved. And the simulation results also verify the effectiveness of the method.

关 键 词：机器人 递归神经网络 回声状态网络 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]