自主水下航行器事件触发自适应神经渐近深度跟踪控制  

Event-triggered adaptive neural asymptotic depth tracking control of autonomous underwater vehicles

作　　者：唐思星 邓英杰 赵运利 TANG Sixing;DENG Yingjie;ZHAO Yunli(Department of Mechanical Engineering,Yanshan University,Qinhuangdao 066004,China)

机构地区：[1]燕山大学机械工程学院,河北秦皇岛066004

出　　处：《大连海事大学学报》2023年第4期57-64,共8页Journal of Dalian Maritime University

基　　金：国家自然科学基金资助项目(52101375);河北省自然科学基金资助项目(E2021203142)。

摘　　要：针对自主水下航行器(AUVs)在未知模型动力学和环境干扰作用下的高精度深度跟踪控制问题,设计一种事件触发自适应神经渐近跟踪控制器。采用径向基神经网络(RBF NNs)逼近模型中的非线性不确定项,将积分有界函数融入控制律和自适应律中,以实现跟踪误差的渐近收敛;采用最小学习参数(MLPs)技术压缩神经网络权重,构造单参数自适应律;在控制器至执行器通道上应用事件触发机制,构造变量式事件触发条件,避免“Zeno”现象;应用径向基函数的不等式关系,解决“代数环”问题;采用Laypunov直接法和Barbalat引理对闭环系统的稳定性进行分析,证明跟踪误差的渐近收敛。仿真试验验证本文控制策略可实现高精度深度跟踪。Aiming at high-precision depth tracking control problem of autonomous underwater vehicles(AUVs)in the unknown model dynamics and environmental disturbances,an event-triggered adaptive neural asymptotic tracking controller was designed.RBF neural networks(NNs)was used to approximate the nonlinear uncertain terms,and the integralbounded functions were incorporated into control laws and adaptive laws to achieve asymptotic convergence of tracking errors.The minimum learning parameters(MLPs)technique was adopted to compress neural weights and construct a single parameter adaptive law.The variable type event triggering condition was constructed by using the event-triggered mechanism on the controller-to-actuator channel,and avoid“Zeno”phenomenon.The inequality relationship of radial basis functions was used to solve the problem of“algebraic loop”.The Lyapunov direct method and Barbarat lemma were used to analyze the stability of the closed-loop system and prove the asymptotic convergence of the tracking error.Simulation experiments verified that the control strategy proposed in this paper has high-precision deep tracking performance.

关 键 词：自主水下航行器(AUVs) 深度跟踪控制 渐近跟踪控制 最小学习参数(MLPs) 事件触发控制(ETC) 

分 类 号：TB712[一般工业技术—真空技术] TP273[自动化与计算机技术—检测技术与自动化装置]