摘要
针对滑模控制具有抖振的缺点,提出了一种基于自组织神经网络的全局滑模控制方法;该方法首先设计了一种全局滑模面,使系统的轨迹一开始就在滑模面上,消除了传统滑模控制的趋近阶段,使系统从初始状态到平衡点的全过程都具有鲁棒性;其次,针对全局滑模变结构控制器系统存在抖振的缺点,利用自组织神经网络的在线学习,自适应地补偿建模误差和外界干扰,削弱了滑模控制的抖振;自组织神经网络能够根据控制目标自适应调节其结构和参数,提高了系统的控制精度;将该方法应用到机械臂控制中,仿真结果表明了其有效性。
A global sliding mode controller based on self--organized neural networks is given for tracking control of manipulators. This method designs a global sliding mode manifold at first, which can make the controlled system in sliding mode at start and eliminates reaching mode phase of normal sliding mode control. Therefore, robustness exists over all the system process. Then, Self--organized neural networks are applied to learn the maximum limit of model errors and disturbances of the controlled system. The controller chattering is weakened. Ac- cording to the command of sliding mode control, the self--organized neural network can adaptively adjust itself structure and parameters and the controller has more control accuracy. The designed controller is applied in manipulator systems and simulation results verify the validity of the control scheme.
出处
《计算机测量与控制》
北大核心
2013年第8期2122-2124,2128,共4页
Computer Measurement &Control
基金
广东省自然科学基金(10451802904006328)
关键词
全局滑模控制
神经网络
建模误差
抖振
滑模面
global sliding mode control
neural network
model error
chattering
sliding mode manifold
