改进的关节机器人神经网络PID控制器  

Improved Neural Network PID Controller of Joint Robots

作　　者：李楠[1] 李文鑫[1] 薛方正[1] 

机构地区：[1]重庆大学自动化学院,重庆400044

出　　处：《控制工程》2013年第6期1052-1054,1059,共4页Control Engineering of China

基　　金：重庆市自然科学基金(2011BB0081);国家自然科学基金(60905053)

摘　　要：针对含有建模误差和不确定干扰的关节机器人轨迹快速跟踪控制,提出了一种改进神经网络PID控制器的设计方法。该方法采用了双控制器鲁棒控制,神经网络通过学习PID的输入输出特性,快速补偿关节机器人系统的建模误差和不确定干扰,而利用最小二乘法和收敛后的神经网络输入输出特性优化PID控制参数,能够削弱建模误差对控制效果的干扰。控制器在李雅普诺夫意义下是稳定的。以两关节机械臂为被控对象进行了仿真实验,实验结果表明改进控制器的优越性。Aiming at the rapid trajectory tracking control with the modeling error and uncertain disturbance problem of the joint robot, this paper presents a design of the improved neural networkPID controller. The architecture employs dualcontroller mode. Through studying the PID controller＇ s input and output characteristics the neural network can compensate the modeling errors and uncertain dis turbance of the joint robot rapidly, and using the LSM and the input and output characteristics of the converged neural network optimize the control parameters of PID controller can weaken the effect of modeling errors on the trajectory tracking control. The controller is sta ble in the Lyapunov sense. In the paper, we have a simulation experiment taking the two linked manipulator as the plant. Finally, the superiority of this design is demonstrated through the experimental results.

关 键 词：关节机器人 神经网络 PID控制 轨迹跟踪 

分 类 号：TP27[自动化与计算机技术—检测技术与自动化装置]