基于模糊补偿的RBF神经网络机械手控制

针对机械手系统的高精度轨迹跟踪控制，提出了一种基于模糊补偿的RBF（radial basis function）神经网络机械手控制方法.该方法首先利用PD（proportional-integral）控制器获得机械手的控制策略，将其输出作为RBF神经网络的输入，并学习得到系统模型；然后运用模糊逻辑补偿器对系统扰动和建模误差进行补偿；最后，在MATLAB/Simulink平台上针对两关节机械臂，进行了有模糊补偿和无模糊补偿系统跟踪的均方根误差测量仿真实验.研究结果表明，两关节机械臂的控制精度分别提高了60.8%和71.4%，本文提出的方法能够解决机械手实际模型很难精确建立的问题，并能对系统未建模部分和扰动部分进行自适应补偿. 

RBF Neural Network Robot Manipulator Control Based on Fuzzy Compensation

For achieving high precision trajectory tracking control of robot manipulators, a control strategy based on fuzzy logic compensation for radial basis function (RBF) neural networks has been proposed. First, the output of a proportional-integral (PD) controller was used in conjunction with an RBF neural network for obtaining a dynamic model of the robot manipulator system. A fuzzy compensator was then introduced for addressing the modelling errors and external disturbances. Furthermore, the fuzzy compensator control scheme and the non-fuzzy compensator control scheme were applied to two-degrees-of-freedom robot manipulators through simulation with MATLAB and Simulink, and the root mean square deviation of tracking errors was thereby measured. The obtained results show that the control accuracies of the first and second joints of the robot manipulators can be improved by 60.8% and 71.4%, respectively. Furthermore, the proposed control scheme can be applied to such robot manipulators which cannot be precisely modelled by compensating the non-modelled part and external disturbances of the system.