机械手臂基于神经网络动态补偿的自适应控制

Adaptive control for manipulator based on neural-network dynamic compensation

研究了模型具有不确定性的机械手臂的跟踪控制问题 .由于模型不确定性的存在 ,基于精确模型设计的控制律很难达到理想的控制效果 .针对这种情况 ,在基于标称模型设计的控制律的基础上 ,采用神经网络来补偿模型的不确定性 ,由于神经网络存在逼近误差 ,因此在控制器设计时 ,引入了H∞ 鲁棒项 ,使得网络逼近误差达到指定的削弱水平并且跟踪误差渐近收敛到零 ,仿真结果表明了该方法的有效性 .

The tracking control of manipulator with model uncertainties and external disturbance is studied. Due to the uncertainties, the controller design based on exact model is difficult to achieve. Thus a neural network is introduced to compensate the uncertainties based on the controller with exact model.Considering the existence of approximation error of the neural network, the H\-∞ robust controller is introduced to reduce the approximation error to a prescribed level and the tracking error tends to zero. The effectiveness of this approach is demonstrated by the simulation examples.