应用神经网络的二自由度欠驱动机械臂控制  

Use Neural Network to Control 2-DOF Underactuated Manipulators

作　　者：朱齐丹[1] 林壮 蔡成涛[1] 

机构地区：[1]哈尔滨工程大学自动化学院,哈尔滨150001

出　　处：《弹箭与制导学报》2006年第S2期377-380,396,共5页Journal of Projectiles,Rockets,Missiles and Guidance

摘　　要：文中通过对水平二自由度欠驱动机械臂的稳态周期运动进行理论分析与计算,发现对于确定的第二个关节无驱动的机器臂系统,如果对其系统中的主动关节施加合适的周期振荡信号,则被动关节会出现一种与输入振幅和频率有关的谐振运动状态。以对驱动臂施加扰动可以将非驱动臂稳定到±π/2这一理论为基础,提出了一种对于具有非驱动关节机器人的基于神经网络的位置开环控制方法,并且对其进行了仿真分析。仿真结果表明,控制效果达到了理想的要求。We present a new control method for two link underactuated manipulator,with a passive rotational second joint.We analyze and calculate the stable periods of systems and found that if we set a suitable periodic oscillation input signal input to the system then the underactuated planar will enter into an oscillation state which correspond with the in- put's frequency and its scope of swings.We follow the theory which the system can be stabilized to the±π/2 through the periodic disturbance of the actuated arm and design an open loop neural network controller which can achieve ideal position control of the underactuated planar.The simulation results show that the control method is effective.

关 键 词：机器人 非驱动关节 非完全约束 神经网络 

分 类 号：TP241[自动化与计算机技术—检测技术与自动化装置]