鲁棒辨识研究的现状

综述了鲁棒辨识问题的研究进展，介绍 了鲁棒辨识问题产生的背景，对各种不同的辨识方法进行了评述，并指出各自的特点，最后总结了今后的发展方向。     

互质因子摄动模型集的闭环l1鲁棒辨识

鲁棒辨识问题已经逐渐成为国际控制界的一个研究热点，本文首先研究了互质因子摄动模型集的闭环ｌ＾１鲁棒辨识问题，建立了系统的Ａ图象拓扑，提出了ＢＩＢＯ不稳定系统的闭环ｌ＾１鲁棒辨识方法，由该方法得到的结果是具体代表性的互质因子摄动模型集。     

基于动态神经网的非线性鲁棒辨识

研究了一类基于动态神经网络的未知非线性多变量系统的鲁棒辨识问题，用Ｌｙａｐｕｎｏｖ稳定性理论获得了具有保护策略的鲁棒调权律，从理论上证明了被辨识的系统是鲁榛 ，辨识误差按建模误差和未建模动态收敛到一个稳定区域，该策略的特点是不需要离线学习又不需要对象的状态完全可测，仿真结果验证了提出的动态网鲁棒辨识策略的有效性。     

基于Volterra泛函级数的非线性系统的鲁棒辨识

针对弱非线性系统的鲁棒建模问题, 基于Volterra泛函级数, 结合集员辨识理论, 提出了广义频率响应函数的鲁棒辨识方法, 形成了一套较完整的弱非线性系统的鲁棒建模方法, 仿真结果表明该方法是行之有效的.     

不确定性系统的鲁棒辨识及其发展

对适合控制的辨识即鲁棒辨识的发展作了综述。首先给出鲁棒辨识的缘由、提法和一些基本概念；然后讨论了集员辨识的理论和发展，Ｈ∞辨识以及部分结论；最后论述了适合控制的辨识理论的进一步发展。     

l1鲁棒辨识:最小二乘算法及试验设计

现有的l^1鲁棒辨识方法依赖于观测数据窗的起始时刻因而不能用来辨识时变系统,针对该问题基于最小二乘法提出了一种l^1鲁棒辨识算法．该算法与观测窗的起始时刻无关,可用于时变系统的辨识．证明了当试验输入为持续激励信号时所提出的算法为本质最优算法,进一步证明了周期持续激励序列为最优试验信号,并给出了辨识误差紧界的计算公式．最后利用提出的算法研究了慢时变系统的l^1鲁棒辨识问题．     

最坏情况下l^1鲁棒辨识方法研究

系统地讨论了ＳＩＳＯ、线性时不变、指数稳定系统在最坏情况下的ｌ＾１鲁棒辨识问题。提出了系统模型集合的最小外框概念，建立了两种任意非零信号作用下ｌ＾１鲁棒辨识算法；提出了任意非零信号作用下系统的可辨识条件；证明了算法的全局收敛性和最优性。     

频域数据非均匀采样时的H_∞鲁棒辨识

考虑了未知系统数据样本在复平面单位圆周上任意采样时的鲁棒辨识问题 ,围绕着充分利用快速傅立叶算法及更有效的利用验后信息 ,提出了一种有效的数据处理方式 ,获得了辨识算法的鲁棒收敛性     

一类非线性时滞系统的H∞模糊控制器的设计——LMI方法

考虑了未知系统数据样本在复平面单位圆周上任意采样时的鲁棒辨识问题,围绕着充分利用快速傅立叶算法及更有效的利用验后信息,提出了一种有效的数据处理方式,获得了辨识算法的鲁棒收敛性.     

一种鲁棒的总体最小二乘自适应辨识算法

针对输入输出观测数据均含有噪声的系统辨识问题，提出了一种鲁棒的总体最小二乘自适应辨识算法．该算法在对总体最小二乘问题与向量的瑞利商及其性质研究的基础上，以被辨识系统的增广权向量的瑞利商(RQ)作为损失函数，利用梯度最陡下降原理导出权向量的自适应迭代算法，并利用随机离散学习规律对权向量模的分析修正了算法梯度，提高了算法的噪声鲁棒性，构成了一种噪声鲁棒的总体最小二乘自适应辨识算法．文中研究了该算法的收敛性能．仿真实验结果表明该算法的鲁棒抗噪性能和稳态收敛精度明显高于其它同类方法，而且可使用较大的学习因子，在较高的噪声环境下仍然保持良好的收敛性．     

面向鲁棒控制的l1中心估计递推算法

讨论未知但有界误差假设下的ｌ１中心估计问题，提出了中心估计递推算法的理论基础，并据此设计了ｌ１中心估计的递推算法。理论分析和数据仿真表明，与原有算法相比，递推算法可以显著减少计算量，提高估计速度，因而可应用于实时在线辨识。所给算法和结果是面向鲁棒控制的。     

Robust performance optimization based on stochastic model errors: the stable case

Robust control aims to account for model uncertainty in design. Traditional methods for robust control typically assume knowledge of hard bounds on the system frequency response. However, this does not match well with system identification procedures which typically yield statistical confidence bounds on the estimated model. This paper explores a new procedure for obtaining a better match between robust control and system identification by using stochastic confidence bounds for robust control design. Given a nominal design, we set up an optimization problem which is aimed at reducing the statistical variability, measured in a mean square sense, from the nominal sensitivity. The proposed procedure is straightforward and leads to an easily computable solution for the final robust controller in the case of a stable plant and modest plant uncertainty. An illustrative example is provided which shows the advantages of the method. Copyright © 2002 John Wiley & Sons, Ltd.     

结构主动控制系统的鲁棒策略

针对目前大多数结构主动控制算法没有考虑结构不确定因素的特点,本文以主动质量阻尼器(AMD:activemass damper)Benchmark结构试验系统为研究对象,提出适合于工程应用的、基于H∞控制理论的主动控制方法.文中建立了主动控制结构的试验系统,根据系统辨识方法得到的面向控制的数学模型,设计YAMD主动控制系统的反馈连接结构,同时对H∞控制权函数的选取以及控制器的设计方法进行了详细的阐述;最后通过试验证明了H∞控制器的有效性和鲁棒性.     

Design and validation of linear robust controllers for nonlinear plants

This paper describes a robust nonlinear control system design procedure inspired by the nonlinear control ideas of Horowitz's Quantitative Feedback Theory. The central concept is the identification of a family of linear time-invariant (LTI) plants that is equivalent to an uncertain nonlinear (and/or time varying) plant in the sense that an LTI controller feasible for this linear plant family is also feasible for the original nonlinear plant. We identify two conditions for evaluating an equivalent linear family (the equivalence condition and the continuity condition) and show that when these two conditions are satisfied an LTI controller that provides satisfactory robust control of an equivalent linear plant family also provides satisfactory robust control for the related uncertain nonlinear plant, independent of the robust design technique used. We then use these two conditions to analyse the validity of the nonlinear QFT design technique published earlier. Our results suggest that nonlinear QFT can be an attractive approach to nonlinear robust control but its validity (in the sense that the linear design solves the nonlinear control problem) can be demonstrated only if additional conditions and contraints not previously reported are satisfied. © 1998 John Wiley & Sons, Ltd.     

含有不确定因素的模型检验及其非伪概率估计*

本文研究了在控制对象的模型具有不确定的建模误差时的模型检验问题，并给邮了控制对象的模型不能被实验数据所否定的概率的估计。     

具有鲁棒调节性能的H-鲁棒次优控制系统设计

本文研究的问题如下，对于具有构造型不确定性的被控对象，设计一动态反馈控制器，使得闭环系统满足Ｈ∞鲁棒次优性能要求。同时，对于满足给定微分方程的参考输入信号，具有鲁棒调节性能。研究结果表明，通过解一适当的Ｈ∞标准设计问题可以得到理想的控制器。基于黎卡提不等式的正定解，给出了控制器的具体设计方法。最后，还给出了该设计方法在二自由度机械手上应用的实验结果。     

基于鲁棒优化的系统辨识算法研究

输入-输出数据是解决系统辨识问题的关键要素,传统的辨识理论除了假定影响输入-输出数据干扰的密度函数已知外,还要假定输入-输出数据能够精确获得,完全忽略了所用数据的质量.本文突破了传统理论的两个假设,首先用工程上易于获得的干扰的有界集合代替干扰的密度函数,并在特定数据不确定性结构下,考虑了数据质量问题,然后,以半定规划为基础,导出了鲁棒对等式,从而将系统辨识转化为对数据质量具有鲁棒性的优化问题,通过求解该优化问题,得到了一种新的鲁棒优化辨识方法,仿真结果表明了新方法的可行性和有效性.     

A robust PI passivity‐based control of nonlinear systems and its application to temperature regulation

This paper deals with the problem of control of partially known nonlinear systems, which have an open‐loop stable equilibrium, but we would like to add a PI controller to regulate its behavior around another operating point. Our main contribution is the identification of a class of systems for which a globally stable PI can be designed knowing only the systems input matrix and measuring only the actuated coordinates. The construction of the PI is carried out invoking passivity theory. The difficulties encountered in the design of adaptive PI controllers with the existing theoretical tools are also discussed. As an illustration of the theory, we consider a class of thermal processes. Copyright © 2015 John Wiley & Sons, Ltd.     

A computer-aided design tool for robustness analysis and control-relevant identification of Horizon Predictive Control with application to a binary distillation column

This paper describes a MATLAB-based computer-aided design tool, IRA-HPC, which accomplishes integrated system identification and robustness analysis for Horizon Predictive Control (HPC), a model predictive control algorithm implemented on the Application Module of the Honeywell TDC 3000 distributed control system. The tool addresses lifecycle as well as functional aspects of the technology, with the goal of making advanced control principles more accessible to the practising control engineer. IRA-HPC systematically performs the various stages of system identification in a control-relevant framework (addressing input design, parameter estimation, and model validation from the standpoint of the final purpose of the model, which is control system design), followed by robust HPC controller tuning using the Structured Singular Value (μ) paradigm as a basis. The benefits of the tool are shown experimentally in the modelling and control of a methanol/isopropanol pilot-scale distillation column, interfaced to an industrial-scale real-time computing testbed. The example demonstrates the practical feasibility of this tool and its benefits in terms of simplifying the choices of design variables in integrated identification and control design.