机械臂关节空间轨迹的神经网络滑模跟踪控制

Manipulator joint space trajectory tracking control based on neutral network-sliding mode controller

为了提高机械臂对给定轨迹的跟踪精度且削弱滑模控制抖振问题,提出了基于RBF神经网络滑模控制的轨迹跟踪方法。建立了多连杆机械臂系统的运动学和动力学模型。首先忽略由建模误差和系统扰动产生的系统不确定项,建立了全局PID滑模控制器,设计了由等效控制律和切换控制律组成的全局滑模控制律;而后使用单隐含层RBF神经网络逼近系统不确定项,使用神经网络对不确定项的逼近值补偿建模误差和系统扰动,达到提高控制精度的目的。经仿真验证,在机械臂初始位置误差较大的情况下,神经网络滑模控制器的调节时间、超调量、驱动力矩抖振远小于全局PID滑模控制器,证明了神经网络滑模控制器在机械臂轨迹跟踪控制中的有效性。

To improve tracking accuracy of manipulator to given trajectory and reduce chattering of sliding control,trajectory tracking method based on RBF neutral network-sliding control is proposed. Kinematics equation and dynamic equation of multiple joints manipulator system are built. Firstly,system uncertain item consisting of model error and system disturbance are neglected,and global PID sliding mode controller is built. Besides,global sliding mode control law consisting of equivalent control law and switching control law is designed. Single hidden layer RBF neutral network is used to approach system uncertain item,and approaching value is used to compensate model error and system disturbance to improve tracking accuracy. Clarified by simulation,in the case of big initial error,setting time,overshoot and driving moment of neutral network-sliding mode controller are far less than global PID sliding mode controller,which can prove validity of neutral network-sliding mode controller on manipulator trajectory tracking control.