不确定下三角非线性系统自适应控制及应用

发布时间：2018-06-17 08:36

本文选题：不确定非线性系统 + 后推方法　；参考：《大连海事大学》2016年博士论文

【摘要】：近年来,基于在线逼近的不确定下三角非线性系统自适应控制吸引了广大学者的研究兴趣。本文对不确定下三角型非线性系统的自适应神经网络控制设计及其在自主式水下机器人的三维轨迹跟踪控制应用方面进行了研究,主要研究工作如下：1.针对一类增益已知的不确定下三角型非线性系统,提出一种基于指令滤波技术的自适应神经网络控制方法。该方法通过一个二阶滤波器而非直接解析地对虚拟控制律求导,从而显著地简化了后推控制器的设计过程,避免了“计算爆炸”问题。此外,将神经网络理想权值的范数做为在线估计参数,使在线学习参数的个数显著地减少,降低了控制器的计算负担。这两种控制技术的有机结合,使得算法结构简单、计算量小而易于工程实现。对系统进行了基于李雅普诺夫稳定性理论的稳定性分析,表明系统中的所有信号是半全局一致最终有界的,且可以通过调节设计参数使得跟踪误差尽可能的小。2.针对一类增益未知的不确定下三角型非线性系统,进行自适应神经网络控制器设计。首先采用动态面控制技术以解决传统后推方法中存在的“计算爆炸”问题,然后将控制器设计过程中出现的未知部分保留到下一步,依此类推,直到控制设计的最后一步,只使用一个神经网络逼近器实现实际控制律中的未知部分,最后将神经网络权值的范数做为在线估计参数,使在线学习参数的个数显著地减少。该方法所设计的控制器结构简单,使计算负担大为减少。应用李雅普诺夫稳定性理论对闭环系统进行分析,得出系统中所有信号半全局一致最终有界的结论,且可以通过调节设计参数使得跟踪误差尽可能的小。3.针对模型未知的自主式水下机器人非线性动力学模型,考虑其存在不确定部分和受到外界干扰的作用,提出两种自适应神经网络控制器设计方法。分别采用指令滤波技术和动态面控制技术来避免传统后推方法中存在的“计算爆炸”问题,然后结合“最少学习参数”的设计思想,使在线学习参数的个数大为减少。所提出的控制算法具有结构简单、计算量小、易于工程实现等特点。对系统进行了基于李雅普诺夫稳定性理论的稳定性分析,表明系统中的所有信号是半全局一致最终有界的,且可以通过调节设计参数使得跟踪误差尽可能的小。最后,利用MATLAB进行了数值仿真研究,仿真结果表明所设计的控制器可以实现自主式水下机器人的三维轨迹精确跟踪控制。
[Abstract]:In recent years, adaptive control of uncertain lower triangular nonlinear systems based on online approximation has attracted many scholars' interest. In this paper, the design of adaptive neural network control for uncertain lower triangular nonlinear system and its application in 3D trajectory tracking control of autonomous underwater vehicle are studied. The main research work is as follows: 1. An adaptive neural network control method based on instruction filtering is proposed for a class of uncertain lower triangular nonlinear systems with known gain. The method uses a second-order filter instead of a direct analytical derivation to the virtual control law, which simplifies the design process of the backstepping controller and avoids the problem of "computational explosion". In addition, the norm of the ideal weight of neural network is taken as the on-line estimation parameter, so that the number of on-line learning parameters is significantly reduced and the computational burden of the controller is reduced. The combination of these two control techniques makes the algorithm simple and easy to implement. The stability analysis of the system based on Lyapunov stability theory shows that all the signals in the system are semi-global uniform and ultimately bounded, and the tracking error can be as small as possible by adjusting the design parameters. An adaptive neural network controller is designed for a class of uncertain lower triangular nonlinear systems with unknown gain. Firstly, the dynamic surface control technique is used to solve the problem of "computational explosion" in the traditional backstepping method, and then the unknown parts in the controller design are retained to the next step, and so on, until the last step of the control design. Only one neural network approximator is used to realize the unknown part of the actual control law. Finally, the norm of neural network weights is taken as the on-line estimation parameter, so that the number of on-line learning parameters is significantly reduced. The controller designed by this method is simple in structure and greatly reduces the computational burden. By using Lyapunov stability theory, the closed-loop system is analyzed, and the conclusion that all the signals in the system are semi-global uniform and finally bounded is obtained, and the tracking error can be as small as possible by adjusting the design parameters. Aiming at the nonlinear dynamic model of autonomous underwater vehicle (AUV) with unknown model, two adaptive neural network controller design methods are proposed, considering the uncertain part of the model and the effect of external disturbance. Instruction filter technique and dynamic surface control technique are adopted to avoid the problem of "computational explosion" in traditional backstepping method, and then the number of online learning parameters is greatly reduced by combining the design idea of "minimum learning parameter". The proposed control algorithm has the advantages of simple structure, small computation and easy engineering implementation. The stability analysis of the system based on Lyapunov stability theory shows that all the signals in the system are semi-global uniform and ultimately bounded, and the tracking error can be minimized by adjusting the design parameters. Finally, the numerical simulation is carried out with MATLAB, and the simulation results show that the controller can realize the accurate tracking control of the autonomous underwater vehicle.
【学位授予单位】：大连海事大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TP273.2

【参考文献】