基于多观测器的UVMS自适应神经网络抗扰控制  

Multiple Observer Based Adaptive Neural Network Anti-disturbance Control of UVMS

作　　者：张宁 王红都[1] 黎明[1] 侯冬冬 ZHANG Ning;WANG Hong-du;LI Ming;HOU Dong-dong(College of Engineering,Ocean University of China,Qingdao 266100,China;713th Research Institute of China State Shipbuilding Corporation,Zhengzhou 450015,China)

机构地区：[1]中国海洋大学工程学院,山东青岛266100 [2]中国船舶集团有限公司713研究所,河南郑州450015

出　　处：《控制工程》2021年第11期2143-2152,共10页Control Engineering of China

基　　金：中央高校基本科研业务费专项(201964012);山东省自然科学基金(ZR202103010892);河南省水下智能装备重点实验室开放基金(20210185)。

摘　　要：水下机器人机械臂系统(UVMS)往往面临高度非线性、模型不确定、复杂海洋干扰及执行器死区非线性等因素,严重影响控制精度。首先将UVMS水动力学模型与死区非线性特性变换为模型已知和未知两部分,其中未知部分与其他干扰归结为总干扰。接下来利用基于径向基函数(RBF)的自适应神经网络逼近未知非线性函数,并构造神经网络性能估计器,利用其估计误差设计了一种新颖的神经网络干扰观测器来估计总干扰。在此基础上提出了一种基于神经网络干扰观测器、性能估计器以及多补偿器的自适应神经网络抗扰控制算法,并利用李雅普诺夫方法分析了闭环系统的稳定性。最后将所提的算法运用到6自由度UVMS进行仿真实验,并通过对比验证了本文所提控制算法的有效性。Underwater vehicle manipulator systems(UVMS) are always subject to difficulties, such as high nonlinearity, model uncertainty, complex ocean disturbances and actuator dead-zone nonlinearity, which seriously affect their control accuracy. Firstly, the UVMS hydrodynamic model and the nonlinear characteristics of dead zone are transformed into the known and unknown parts, in which the unknown part and disturbances is lumped into the total disturbances. Secondly, an adaptive neural network based on radial basis function(RBF) is used to approximate the unknown nonlinear function, and a novel neural network disturbance observer is designed to estimate the total disturbance by using the estimation error from a designed neural network performance estimator. Subsequently, an adaptive neural network anti-disturbance control algorithm is proposed for UVMS based on neural network disturbance observer, performance estimator, and multiple compensators. The stability of the closed-loop system is analyzed via the Lyapunov method. Finally, the proposed algorithm is applied to the 6-degree-of-freedom UVMS for simulation experiments and the effectiveness of the proposed control algorithm proposed in this paper is verified by comparison.

关 键 词：RBF神经网络控制 水下机器人机械臂系统 神经网络性能估计器 神经网络干扰观测器 执行器死区 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]