Computer Engineering and Applications ›› 2022, Vol. 58 ›› Issue (1): 268-273.DOI: 10.3778/j.issn.1002-8331.2007-0366

• Engineering and Applications • Previous Articles     Next Articles

Sliding Mode Convolutional Neural Network Trajectory Tracking Control for Robot Manipulators

XIE Hong, WANG Lichen, YUAN Xiaofang, CHEN Haibin   

  1. College of Electrical and Information Engineering, Hunan University, Changsha 410082, China
  • Online:2022-01-01 Published:2022-01-06

机械臂卷积神经网络滑模轨迹跟踪控制

谢宏,王立宸,袁小芳,陈海滨   

  1. 湖南大学 电气与信息工程学院,长沙 410082

Abstract: In view of the development of industrial technology, there are high requirements for the accuracy and rapid control of the multi-joint manipulator. For this, a sliding mode convolutional neural network trajectory tracking control method for robot manipulators is proposed. Firstly, the paper analyzes the dynamic equation of the robot manipulators and extracts the uncertain part. Secondly, it constructs a deep convolutional neural network to compensate the uncertain part. Finally, the compensation part is introduced into the sliding mode control law, and the precise control of the trajectory tracking of the robot manipulators is realized through the improved sliding mode control, and the stability of the system is demonstrated by constructing a Lyapunov function. The simulation results show that this method can meet the requirements of trajectory tracking and reduce the vibration. Through comparison with the other three typical control methods, the test results show that this method can accelerate the convergence of the trajectory tracking errors, and the tracking accuracy has been significantly improved.

Key words: convolutional neural network, sliding mode control, trajectory tracking, deep learning

摘要: 针对工业技术的发展对于多关节机械臂的精度与快速控制高要求,提出了一种机械臂卷积神经网络滑模轨迹跟踪控制方法。分析机械臂动力学方程,提取其中的不确定部分,针对不确定部分,构建深度卷积神经网络对其进行补偿,将补偿部分引入到滑模控制律中,通过改进后的滑模控制实现对机械臂轨迹跟踪的精确控制,并通过构建Lyapunov函数论证了系统的稳定性。仿真结果显示该方法能够满足轨迹跟踪要求,且减小了抖振现象。通过与其余三种典型控制方法的对比,测试结果表明,该方法加快了轨迹跟踪误差的收敛,且跟踪精度有了明显的提高。

关键词: 卷积神经网络, 滑模控制, 轨迹跟踪, 深度学习