中国机械工程

• 机械基础工程 • 上一篇    下一篇

无模型机械臂BP神经网络状态观测及反演跟踪控制

李光;符浩   

  1. 湖南工业大学,株洲,412007
  • 出版日期:2016-04-10 发布日期:2016-04-11
  • 基金资助:
    国家自然科学基金资助项目(61273157)

BP  Neural Network State  Observation  and  Backstepping  TrackingControl of  Model-free  Robotic  Manipulators

Li Guang;Fu Hao   

  1. Hunan University  of  Technology,Zhuzhou,Hunan,412007
  • Online:2016-04-10 Published:2016-04-11
  • Supported by:

摘要: 针对摩擦阻尼及模型参数不确定的情况,运用反演控制设计策略,针对多连杆机械臂提出了一种基于神经网络观测器的无模型轨迹跟踪控制方法。运用带有修正项的自适应BP神经网络观测器对不可测状态量进行观测,同时对系统模型进行在线逼近。在此基础上设计了基于观测状态和逼近模型的反演跟踪控制器, Lyapunov稳定性理论证明了该控制器能够保证跟踪误差的有界和闭环系统中所有信号的有界。跟踪给定轨迹的仿真实验证明了该方法的有效性。

关键词: 机械臂, 状态观测器, BP神经网络, 反演控制

Abstract: A  model-free  trajectory tracking control  algorithm was proposed for multi-link robotic manipulators with friction damping and model parameter uncertainties based on the backsteeping techniques. An adaptive BP neural network state observer with modification items was used  to obtain the unmeasured states and approximated the system model online.Then,a backstepping tracking controller was developed, which took advantages of the observed state values and dynamics.Based on Lyapunov stability theory it is proved that  presented controller can guarantee the tracking accuracy and all signals involved are bounded. Finally, it demonstrates that the strategy is effective by the simulation of tracking expected trajectory.

Key words: robotic , manipulator, state observer, BP neural network, backstepping control

中图分类号: