基于神经网络的机器人轨迹跟踪控制  

Trajectory Tracking Control of Robot Manipulators Using Neural Networks

作　　者：任雪梅[1] 

机构地区：[1]北京理工大学自动控制系,100081

出　　处：《控制与决策》1997年第4期317-321,384,共6页Control and Decision

摘　　要：针对机器人模型未知情况,讨论了用神经网络和反馈控制实现机械手的跟踪控制。提出一种基于参考误差的投影算法来训练网络权值,训练后网络输出能逼近期望的前馈力矩,并从理论上证明跟踪误差的收敛性。仿真结果表明该方案具有较好的跟踪性能和较强的抗干扰能力。A new controller is discussed for robot manipulators with no exact dynamics, by means of neural networks and feedback. The neural network trained by projection algorithm with the reference error is used to approximate the desired feedforward compensation. Convergence of the tracking error is proved. Simula-tion results show that the proposed scheme has more satisfactory tracking performance and better robustness for disturbances than PD feedback controllers.

关 键 词：机器人 神经网络 轨迹跟踪控制 

分 类 号：TP242.6[自动化与计算机技术—检测技术与自动化装置]