SIA OpenIR  > 机器人学研究室
非线性机电系统的鲁棒及预测控制
Alternative TitleRobust and Predictive Control of Nonlinear Mechatronic Systems
何玉庆1,2
Department机器人学研究室
Thesis Advisor韩建达
ClassificationTH-39/
Keyword非线性机电系统 扰动抑制 非线性鲁棒控制 非线性预测控制 控制lyapunov函数
Call NumberTH-39/H33/2008
Pages154页
Degree Discipline模式识别与智能系统
Degree Name博士
2008-02-19
Degree Grantor中国科学院沈阳自动化研究所
Place of Conferral沈阳
Abstract近年来,机器人、数控机床等机电系统在国民生产及生活中得到了越来越广泛的应用,与之相对应的,机电系统的控制无形中也逐渐成为机电一体化和自动控制的研究热点。另一方面,非线性特性是任何实际系统普遍存在的现象,在机电系统中尤其如此。本文从控制器设计的角度研究非线性机电系统的两个典型问题:系统闭环优化性能的改善;控制器鲁棒性的增强。这也是当前自动控制研究领域的两个重点及热点问题。 最优性是闭环系统性能最实用的评价指标之一。最优控制及预测控制是试图实现控制性能优化的两种典型方法。但前者因本身不构成闭环而严重缺乏鲁棒性,在实践中很少能得到应用;后者作为前者理念的推广,在实践中已经得到了较为广泛地应用,并取得了不错的效果,但在实现某种程度的鲁棒性时依然面临困难。除最优性外,闭环控制的鲁棒性也是非线性系统控制中亟待解决的问题之一,这是由于实际的系统几乎不可能避免模型不确定性。 现有的非线性预测控制及鲁棒控制方法无论在方法的广泛适用性还是在可行性方面都还远非完备。据此,本论文沿鲁棒性和最优性两条主线,以典型的机电系统(无人直升机模型)为研究对象,分别进行了深入的理论和实验研究,并最终形成一种同时兼顾最优性和鲁棒性的控制器设计框架,以期在一定程度上解决现有方法中存在的问题,并为以后更深入的研究工作奠定基础。 鉴于此,本论文分别针对鲁棒控制和预测控制展开讨论,其中前者主要解决基于加速度反馈实现鲁棒控制的方法,内容为第二章和第三章;后者则旨在解决基于控制Lyapunov函数方法实现实时稳定预测控制,主要内容为第四章和第五章。本论文的具体内容安排如下: 论文的第一章综述了控制理论在鲁棒性与最优性两个方向的发展概况(主要针对非线性系统),包括其发展历史,现存方法的局限性等。从而引出本论文的研究内容及研究意义。 第二章,研究了基于加速度反馈的控制器鲁棒增强方法。在深入分析常规加速度反馈控制方法的基础上,指出其存在的三方面主要问题:代数环问题;高增益实现问题和不能用于欠驱动非线性系统等。并针对两种典型的非线性系统(以无人直升机模型为代表)将新的加速度反馈控制方法与H∞控制相结合,得到了一种能够保证输入输出稳定的扰动抑制方法。大量的仿真结果验证了方法的可行性及有效性。 随后,在第三章研究了加速度的估计问题。基于加速度反馈的鲁棒控制器增强技术得以实现的前提是加速度信号的获取,本章在分析了现有加速度估计方法存在严重的滞后问题的同时,提出了将Kalman滤波方法同牛顿预测方法相结合以改善相位滞后问题的方法。实验及仿真结果验证了方法的有效性。 第四章提出了基于控制Lyapunov函数的稳定闭环控制器设计框架。本章利用集值分析理论研究了控制Lyapunov函数具有的一些性质及其在控制器设计中的应用。随后,介绍了两种典型的根据控制Lyapunov函数设计控制器的方法。接着,将引导函数的概念引入到Freeman的逐点最小范数控制方法中,形成了一种新的利用控制Lyapunov函数设计非线性控制器的方法—广义逐点最小范数控制器。最后指出,在这种框架下,鲁棒控制器设计也可以实现,并针对三种不同的不确定性系统给出了鲁棒广义逐点最小范数控制器设计方法。 最后,在第五章,将前面提出的广义逐点最小范数控制引入到非线性预测控制中去,以期利用控制Lyapunov函数保证闭环稳定性,同时利用控制器中的参数化变量作为优化对象以减轻预测控制算法的计算负担,从而达到实时稳定预测控制的目的。另外,在这一章我们还在第二章和第四章的基础上,结合加速度反馈思想和鲁棒控制Lyapunov函数的概念,提出了一种用于扰动抑制的鲁棒实时预测控制算法。同样,仿真实验验证了方法的有效性和可行性。
Other AbstractIn recent years, more and more mechatronic systems, such as robot, CNC machine tool, etc., are being used. Consequently, the control of mechatronic systems, which tries to achieve higher performance, has been attracting more and more researches. Linear control, for example, has been successfully used in most of the existing mechatronic systems. However, whith the increasing complexity in the system and the demands on its performance, nonlinearity has to be taken into the control design. With respect to nonlinear control, the robustness and optimization may be two of the most important issues. This thesis mainly studies the methodology that can improve the robustness and optimization performance of a nonlinear system. Optimization is one of the most practical evaluation indexes of a closed loop control. Optimal control and predictive control are two methodologies that try to realize the optimization. Optimal control, which is inherently in open loop structure, fails to obtain extensively application in real plants due to the lack of robustness. Model predictive control, on the other hand, is a kind extension of optimal control, and can realize some of the optimal performance, but is is still difficult while trying to achieve robustness simultaneously. Robustness is another important performance index for a control method besides the optimization, especially for the implementation on real system. This is because that there always exist differences between a real plant and its dynamics model. If the control designed according to the model can not reject the differences and behaves poor performance or even loses stability, which will be unacceptable. With respect to the open problems existed in the current control design, methods that can improve the robustness and optimization performance of nonlinear systems are studied in this thesis, under the frame of, respectively, acceleration feedback and Control-Lyapunov-Function enhanced predictive control scheme. Novel control methods are proposed, theoretically proved, and tested mainly with respect to the model of an unmanned helicopter, which presents a typical nonlinear, coupling and underactuated dynamics. Finally, the proposed robustness enhancement and optimization techniques are integrated into the so-called robust-optimal nonlinear predictive control in the end of the thesis. The main contents of this thesis are organized as follows: In Chapter 1, the former and current researches on robust, optimal and model based predictive control are summarized. The problems existing in available methods are sited and analized. The motivation of the studies in this thesis is introduced thereafter. In Chapter 2, a generalized acceleration feedback control (GAFC) scheme is proposed for nonlinear and possible underactuated systems. A pre-filter is introduced to the classical AFC with the purpose of releasing the difficulties existing in the implementation of classical AFC, which include the algebraic loop, the necessity of high acceleration feedback gain, and the applicability on nonlinear underactuated systems. As a robustness enhancement scheme, the proposed method is tested with respect to the dynamics of an unmanned helicopter, and the simulation results demonstrate the feasibility and improvements. Chapter 3 is about the estimation of acceleration signals. In order to realize AFC, the acceleration signal is necessary. The acceleration signal has to be estimated in stead of measuring in the cases that accelerometer implementation is difficult or impossible. A novel acceleration estimator is proposed by integrating Kalman filter and Newton Predictor together. The acceleration estimated by the proposed method is experimentally compared with those obtained by other estimators and also measured by accelerometer. The GAFC using estimated acceleration is also implemented on the helicopter model and the results are compared with those using the ‘real’ acceleration, to show the feasibility of the estimated acceleration to be used in control. In Chapter 4, a new controller design framework based on the concept of Control Lyapunov Function (CLF) is given. The properties of CLF are analyzed by using set-valued analysis technique. Subsequently, two typical controller design methods based on CLF are proposed. Moreover, the main problem existing in the proposed controller, i.e., lack of design parameters is pointed out and analyzed. Based on these, a new controller design framework -- generalized pointwise min-norm controller (GPMN) -- is presented by introducing a guide function into Freeman’s pointwise min-norm controller given in 1996. Finnaly, three robust GPMN controllers are also given based on three different kinds of uncertainty systems. All of these control methods are theoretically proved, and tested with respect to the helicopter dynamics. Finally, in Chapter 5, the real time nonlinear model predictive control (NMPC) with ensured closed loop stability is discussed based on the GPMN controller. The GPMN structure is inserted into the NMPC to ensure the stability, and the parameterized guide function is used to reduce the computational burden of original NMPC algorithms. In the end, the new AFC method is also integrated into the NMPC to further improve its robustness. Extensive simulations are conducted with respect to the helicopter dynamics to show the performance and improvements of the proposed methods.
Language中文
Contribution Rank1
Document Type学位论文
Identifierhttp://ir.sia.cn/handle/173321/290
Collection机器人学研究室
Affiliation1.中国科学院沈阳自动化研究所
2.中国科学院研究生院
Recommended Citation
GB/T 7714
何玉庆. 非线性机电系统的鲁棒及预测控制[D]. 沈阳. 中国科学院沈阳自动化研究所,2008.
Files in This Item:
File Name/Size DocType Version Access License
LWHYQ01.pdf(2741KB) 开放获取--Application Full Text
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[何玉庆]'s Articles
Baidu academic
Similar articles in Baidu academic
[何玉庆]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[何玉庆]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.
 

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.