基于奇异摄动与神经网络的柔性臂控制

运用拉格朗日法建立了臂杆柔性的机械臂的动力学方程。为解决柔性臂末端位置的跟踪及克服柔性臂在运动中的振动问题。运用奇异摄动法将系统分解成快、慢两个子系统，设计混合控制器。通过设计神经网络控制器线性化慢系统。使其轨迹跟踪期望值。用线性状态反馈配置极点稳定快系统，抑制振动。一个单杆柔性机械臂的仿真算例表明，本文的控制方法保证了柔性臂刚性运动的精确跟踪，同时消除了弹性振动，避免了零动力学不稳定问题。

Control of Robot Arms with Link Flexibility Using Singular Perturbation and Neural Network

The dynamic equation of robot arms with link flexibility is established using Lagrange method. To satisfy the tracking requirements of the link-tip position and suppress the oscillation of the link, the singular perturbation approach is used to decompose the robots into a slow subsystem and a fast subsystem, which allows a combined control design for the original system. A neural network controller is designed to linearize the slow subsystem and make the rigid motion track a desired trajectory. Using state-feedback techniques, the poles of the fast subsystem are assigned to stabilize the fast system and attenuate the oscillatory of the link. A simulational example of a single flexible manipulator shows that the combined control is effective in rigid motion tracking control and oscillatory attenuation control, and the unstable zero dynamics can be avoided.