Stewart平台神经网络非奇异终端滑模控制

PDF下载HTML

Title:

Neural network-based nonsingular terminal sliding mode control of the Stewart platform

作者:

常光宇¹, 陈志峰², 郭春雨³, 庞明¹

1. 哈尔滨工程大学 智能科学与工程学院, 黑龙江 哈尔滨 150001;
2. 哈尔滨商业大学 能源与建筑工程学院, 黑龙江 哈尔滨 150028;
3. 哈尔滨工程大学 青岛创新发展基地, 山东 青岛 266000

Author(s):

CHANG Guangyu¹, CHEN Zhifeng², GUO Chunyu³, PANG Ming¹

1. College of Intelligent Systems Science and Engineering, Harbin Engineering University, Harbin 150001, China;
2. College of Energy and Architectural Engineering, Harbin University of Commerce, Harbin 150028, China;
3. Qingdao Innovation and Develo

关键词:

Stewart平台; 并联机器人; 动力学; 滑模控制; 自适应控制系统; 神经网络; Lyapunov方法; 非线性控制

Keywords:

Stewart platform; parallel robot; dynamics; sliding mode control; adaptive control system; neural networks; Lyapunov methods; nonlinear control

分类号:

TP242.2

DOI:

10.11992/tis.202210004

文献标志码:

2023-12-01

摘要:

针对Stewart平台的六自由度（six degrees of freedom, 6-DOF）轨迹跟踪问题，提出一种基于神经网络的非奇异终端滑模控制方法并应用于Stewart平台的位置姿态控制中。通过分析Stewart平台的位置反解和速度反解，建立运动学方程，利用牛顿?欧拉方程建立动力学方程，并结合加速度反解得到了平台的状态空间表达式；基于非奇异滑模面函数，设计非奇异终端滑模控制律。考虑到径向基函数（radial Basis function, RBF）神经网络的逼近特性，采用RBF神经网络对模型未知部分进行自适应逼近，并利用Lyapunov第二法设计了自适应律；通过仿真证明控制器设计的有效性。仿真结果表明，相比于比例积分微分（proportional integral derivative, PID）控制器，提出的RBF神经网络非奇异终端滑模控制器具有更好的轨迹跟踪精度和动态特性。

Abstract:

This paper proposes a solution to the six degrees of freedom trajectory tracking problem of the Stewart platform using a nonsingular terminal sliding mode control method based on a neural network. This method is applied to the position and pose control of the Stewart platform. First, a kinematic equation is established by analyzing the position inverse solution and velocity inverse solution of the Stewart platform. Simultaneously, the dynamic equation is established based on the Newton-Euler equation. By integrating the acceleration inverse solution, we obtain the state–space representation of the platform. Subsequently, a nonsingular terminal sliding mode control law is designed using the nonsingular sliding surface function. Considering the approximation characteristics of the radial basis function (RBF) neural network, we employ this network to adaptively approximate the unknown term of the equation. An adaptive law is then designed based on the second method of Lyapunov. Finally, the effectiveness of the controller design is proved through simulations. The simulation results show that the proposed controller that uses an RBF neural network and nonsingular terminal sliding mode outperforms the proportional integral derivative controller in terms of trajectory tracking accuracy and dynamic characteristics.