数学家传记
鲁道夫·卡尔曼是一位匈牙利出生的美国数学家,以其在控制理论方面的工作而闻名。
鲁道夫·卡尔曼移居美国,在麻省理工学院学习电气工程。他于1953年获得理学学士学位,并继续在该校攻读研究生,于1954年获得理学硕士学位。为了攻读博士学位,他前往纽约市的哥伦比亚大学,师从弗瑞兹·约翰 Ralph Ragazzini。他于1955年被任命为哥伦比亚大学控制论讲师,1957年晋升为兼职助理教授。他于1958年提交了博士论文Analysis and Synthesis of Linear Systems Operating on Randomly Sampled Data,并于同年获得理学博士学位。
在获得博士学位之前,卡尔曼已经开始发表有影响力的论文。例如在Physical and mathematical mechanisms of instability in nonlinear automatic control systems(1959)中,他给出了(引用Horace Trent的评论):-
……对含有一个或多个非线性元件的控制系统的稳定性问题进行了清晰的讨论。其写作风格既吸引高级设计工程师,也吸引研究此类系统理论的研究人员。讨论仅限于那些可以被视为由有限数量的具有时不变参数的“集中”部件组成的系统。因此,这样的系统可以用有限个联立常微分方程来描述。作者利用拓扑学方法来分析非线性系统;即,他考察由输入变化、初始条件变化等在相空间中生成的轨线的性质,并从这些结果中推断出关于结点、焦点和鞍点可能存在的定量信息。在这些分析过程中,他使用了虚拟临界点方法。该论文的主要目标之一是对可能引起不稳定的非线性类型提出分类。
获得博士学位后,卡尔曼被任命为马里兰州巴尔的摩高级研究所的研究数学家。他从1958年到1964年在那里工作,在此期间他产生了一些极其重要的成果。例如,他与J E Bertram合著了三篇论文:A unified approach to the theory of sampling systems(1959);Control system analysis and design via the "second method" of Lyapunov. I. Continuous-time systems(1960);以及Control system analysis and design via the "second method" of Lyapunov. II. Discrete-time systems(1960)。
还有三篇论文A new approach to linear filtering and prediction problems(1960)、Mathematical description of linear dynamical systems(1963)以及(与Richard S Bucy合著)New results in linear filtering and prediction theory(1961)。作者对最后提到的论文总结如下:-
针对最优滤波误差的协方差矩阵,导出了一个雅各布·黎卡提类型的非线性微分方程。这个“方差方程”的解完全确定了有限或无限平滑区间以及平稳或非平稳统计下的最优滤波器。方差方程与变分法中的哈密顿(正则)微分方程密切相关。在某些情况下可以得到解析解。通过一些例子说明了方差方程的意义,这些例子重复、简化或扩展了该领域的早期结果。联系随机估计与确定性控制问题的对偶原理在理论结果的证明中起着重要作用。在几个例子中,估计问题及其对偶问题被并排讨论。方差方程的性质在自适应系统理论中引起了极大的兴趣。对此的一些方面进行了简要考虑。
现在我们将最后提到的这三篇论文联系起来,因为它们“被证明在其领域具有根本性或持久的重要性”,并因此被美国数学会授予1986年凯瑟琳·斯蒂尔奖。引文如下[2]:-
这些论文中提出的思想是系统和控制的现代理论与实践的基石。它们不仅导致了极其有用的发展,例如卡尔曼-Bucy滤波器,而且促进了系统理论的迅速进展,如今系统理论借鉴了从微分方程到代数几何的数学。
卡尔曼没有出席在德克萨斯州圣安东尼奥举行的第93届美国数学会年会来领取该奖,但他在会上宣读了他的回应,试图解释这些论文的重要性。他的回应开头如下[2]:-
我从一开始(1959年1月底,引文中第二篇论文的诞生日期)就意识到,对现在被称为卡尔曼滤波的某些东西的深入分析具有重大意义。但即使有这种不谦虚,我也没有完全预料到这项工作的所有反应。到目前为止,已有大约1000篇相关出版物,至少两篇引文经典,等等。有些事情需要解释。
在THIS LINK阅读卡尔曼1986年对获得斯蒂尔奖的回应更完整的报告。
让我们回到提供卡尔曼职业生涯的细节。他于1964年离开马里兰州巴尔的摩的高级研究研究所,担任斯坦福大学教授,涵盖工程力学、电气工程和数学系统理论。1971年,他转到盖恩斯维尔的佛罗里达大学,在那里他同时担任数学系、电气工程系以及工业与系统工程系的 graduate research professor,并担任数学系统理论中心主任。两年后,在继续担任佛罗里达大学这些职位的同时,他还成为苏黎世联邦理工学院的教授。卡尔曼于1992年从佛罗里达大学的职位退休(当时他成为荣休教授),并于1997年从苏黎世联邦理工学院退休。
我们已经考察了卡尔曼在职业生涯早期所写的一些重要论文。为了更全面地了解他的贡献,我们引用Eduardo Sontag和Yutaka Yamamoto的话,两人都是卡尔曼在佛罗里达大学的博士生(分别于1976年和1978年获得博士学位)[5]:-
在20世纪60年代,卡尔曼是控制系统严格理论发展的领导者。在他众多杰出贡献中,包括对最基本状态空间概念的表述和研究(包括能控性、能观性、最小性、从输入/输出数据的可实现性、矩阵雅各布·黎卡提方程、线性二次控制和分离原理),这些概念如今在控制中无处不在。虽然其中一些概念也在其他背景下遇到,如最优控制理论,但正是卡尔曼认识到它们在系统分析中所起的核心作用。由卡尔曼表述的范式和他建立的基本结果已成为控制和系统理论基础的内在组成部分,并且是研究以及该领域每一部论述中的标准工具,从本科工程教科书到研究生数学研究专著。在20世纪70年代,卡尔曼在将代数和几何技术引入线性和非线性控制系统研究中发挥了重要作用。自20世纪80年代以来,他的工作集中在统计基础、计量经济建模和辨识的系统论方法上,作为他早期对最小性和可实现性研究的自然补充。
除了许多研究论文外,卡尔曼还与Peter L Falb和Michael A Arbib合著了Topics in mathematical system theory(1969年)一书。作者们在序言中写道:-
本书并不自诩为一部系统论著。相反,它旨在呈现当今的数学系统理论——一个活跃、富有挑战性、令人兴奋、困难、混乱、有回报且很大程度上尚未探索的领域。本书的四个部分各自独立。没有试图消除重复;没有标准化符号。
我们上面提到卡尔曼在1986年获得了斯蒂尔奖。这是他因杰出的创新贡献而获得的众多奖项和荣誉之一。除了斯蒂尔奖,他还于1974年获得电气和电子工程师协会荣誉奖章,1984年获得电气和电子工程师协会百年奖章,1985年获得稻盛基金会高科技京都奖,因为:-
……现代控制和系统理论的创造者。卡尔曼理论建立于20世纪60年代初,为控制工程带来了根本性变革,并从此为现代控制理论的迅速发展奠定了基础。
他还在1997年获得了Richard E 理查德·贝尔曼 Control Heritage Award,并在2008年获得了国家工程院的Charles Stark Draper Prize:-
……因开发和传播最优数字技术(称为卡尔曼滤波器),该技术被广泛用于控制大量消费、健康、商业和国防产品。
2009年10月7日,美国总统巴拉克·奥巴马在白宫的一个颁奖仪式上表彰了卡尔曼,向他颁发了国家科学奖章,这是美国对科学成就所能授予的最高荣誉。Eduardo Sontag在[4]中解释了卡尔曼滤波的重要性:-
在卡尔曼的早期工作中,有一项是开发了现在称为卡尔曼滤波器的技术,用于检测噪声中的信号。这通过为滤波问题提供一种递归方法,彻底改变了估计领域。在卡尔曼滤波器出现之前,大多数数学工作都基于诺伯特·维纳的思想,但“诺伯特·维纳滤波”被证明难以应用。卡尔曼的方法基于状态空间技术和递归最小二乘算法,开辟了许多新的理论和实践可能性。卡尔曼滤波对应用数学、工程和科学的所有领域产生了巨大影响。甚至无法开始列举其实际应用。仅作为其多样性的例子,可以提到阿波罗航天器和商用飞机的制导、在地震数据处理、核电站仪器和人口模型中的用途,以及在计量经济学中的应用。
除了上述荣誉外,他还被选为美国国家科学院、美国国家工程院、American Academy of Arts and Sciences、Hungarian Academy of Sciences、Académie des Sciences和苏联科学院的成员。
卡尔曼关于统计学的思想发人深省,自退休以来,他就这一主题撰写了许多有趣的文章。例如,他发表了一个非正式讲座,出版为Probability and science(1993),其中他(引用Zeno G Swijtink的评论):-
……哀叹二战后probability应用的爆炸式增长,并质疑概率是否存在于现实世界中。随机性确实存在,为了更好地刻画它,他提出了一个新定义:随机不是由简单的经典规则唯一决定的。根据这个定义,所有无理数都被认为是随机的,大多数“3x + 1”映射也是如此。作者总结道,混沌理论家和概率论家之间缺乏互动也表明,经典的概型与现实世界是多么不相干。
1995年,他发表了Randomness and probability,2002年发表了关于What Is a Statistical Model?的讨论。他认为[3]:-
……当前被接受的统计模型概念并不科学;相反,它是对可能构成(科学)现实的东西的一种猜测,缺少了至关重要的反馈要素,也就是说,没有用现实来检验所假设的、假定的、期望的、看似自然的(但实际上只是猜测的)模型。坦率地说,就今天所知,并不存在具体的i.i.d.(独立同分布)过程,不是因为这不令人向往、不美好,甚至不优美,而是因为大自然似乎并非如此。(历史旁注:回想一下,物理学家曾一度认为以太是如此必要、不可避免、吸引人、清晰而优美的概念,以至于它必定存在;唉,所有在世的物理学家都不得不认识到,这样的论证无法带来好的科学。)正如伯特兰·罗素在献身哲学的漫长一生结束时所说:“粗略地说,我们所知道的是科学,我们所不知道的是哲学。”在科学语境中,但也许不是在应用领域,我担心今天的统计建模属于哲学的范畴。
最后注意,卡尔曼与Constantina Stavrou结婚,后者在Johns 威廉·霍普金斯大学学习经济学;他们有两个孩子,Andrew和Elisabeth。
Rudolf Kalman emigrated to the United States where he studied electrical engineering at the Massachusetts Institute of Technology. He was awarded a S.B. in 1953, and continued to undertake graduate studies there leading to a S.M. in 1954. For his doctoral work he went to Columbia University in New York City where he was advised by John Ralph Ragazzini. He was appointed as an Instructor in Control Theory at Columbia University in 1955 and promoted to Adjunct Assistant Professor in 1957. He submitted his doctoral thesis Analysis and Synthesis of Linear Systems Operating on Randomly Sampled Data in 1958 and was awarded a D.Sci. in that year.
Before the award of his doctorate Kalman had begun to publish influential papers. For example in Physical and mathematical mechanisms of instability in nonlinear automatic control systems (1959) he gives (quoting from a review by Horace Trent):-
... a lucid discussion of the stability problem for control systems containing one or more nonlinear elements. It is written in a style which appeals equally to advanced design engineers and to research workers who deal with the theory of such systems. The discussion is limited to those systems which can be treated as if they were made up of a finite number of "lumped" parts with time invariant parameters. Thus, such a system can be described by a finite number of simultaneous ordinary differential equations. The author exploits the topological approach to an analysis of non-linear systems; i.e., he examines the nature of trajectories in phase space as generated by variations of inputs, variation of initial conditions, etc., and from these results deduces quantitative information about the possible existence of nodes, foci, and saddle points. During these analyses he makes use of the method of virtual critical points. One of the main goals of the paper is to propose a classification of the types of nonlinearities which can give rise to instabilities.
After the award of his doctorate Kalman was appointed as a Research Mathematician at the Research Institute for Advanced Studies in Baltimore, Maryland. He worked there from 1958 to 1964 and during this time he produced some results of outstanding importance. For example there are three papers he wrote jointly with J E Bertram: A unified approach to the theory of sampling systems (1959); Control system analysis and design via the "second method" of Lyapunov. I. Continuous-time systems (1960); and Control system analysis and design via the "second method" of Lyapunov. II. Discrete-time systems (1960).
There are also the three papers A new approach to linear filtering and prediction problems (1960), Mathematical description of linear dynamical systems (1963) and (with Richard S Bucy) New results in linear filtering and prediction theory (1961). The authors summarise this last mention paper as follows:-
A nonlinear differential equation of the Riccati type is derived for the covariance matrix of the optimal filtering error. The solution of this 'variance equation' completely specifies the optimal filter for either finite or infinite smoothing intervals and stationary or non-stationary statistics. The variance equation is closely related to the Hamiltonian (canonical) differential equations of the calculus of variations. Analytic solutions are available in some cases. The significance of the variance equation is illustrated by examples which duplicate, simplify, or extend earlier results in this field. The duality principle relating stochastic estimation and deterministic control problems plays an important role in the proof of theoretical results. In several examples, the estimation problem and its dual are discussed side-by-side. Properties of the variance equation are of great interest in the theory of adaptive systems. Some aspects of this are considered briefly.
Now we have linked these last-mentioned three papers since they "proved to be of fundamental or lasting importance in their field" and as such were awarded the 1986 Steele Prize by the American Mathematical Society. The citation reads [2]:-
The ideas presented in these papers are a cornerstone of the modern theory and practice of systems and control. Not only have they led to eminently useful developments, such as the Kalman-Bucy filter, but they have contributed to the rapid progress of systems theory, which today draws upon mathematics ranging from differential equations to algebraic geometry.
Kalman was not present at the 93rd American Mathematical Society Annual meeting in San Antonio, Texas, to receive the Prize but his response, attempting to explain the significance of the papers, was read at the meeting. His response begins as follows [2]:-
I have been aware from the outset (end of January 1959, the birthdate of the second paper in the citation) that the deep analysis of something which is now called Kalman filtering were of major importance. But even with this immodesty I did not quite anticipate all the reactions to this work. Up to now there have been some 1000 related publications, at least two Citation Classics, etc. There is something to be explained.
Read a fuller report of Kalman's 1986 response to the award of the Steele Prize at THIS LINK.
Let us return to giving details of Kalman's career. He left the Research Institute for Advanced Studies in Baltimore, Maryland, in 1964 to take up the position of Professor at Stanford University covering engineering mechanics, electrical engineering, and mathematical system theory. In 1971 he moved to the University of Florida, Gainesville, where he was Graduate Research Professor jointly in the Departments of Mathematics, Electrical Engineering, and Industrial and Systems Engineering, and Director of the Center for Mathematical System Theory. Two years later, while continuing to hold these positions at the University of Florida, he also became a Professor at the Eidgenössische Technische Hochschule in Zürich. Kalman retired from his positions at the University of Florida in 1992 (when he became Professor Emeritus) and from the Eidgenössische Technische Hochschule in Zürich in 1997.
We have looked at some of the important papers Kalman wrote in the early part of his career. To get a fuller overview of his contributions we quote from Eduardo Sontag and Yutaka Yamamoto, both former doctoral students of Kalman at Florida University (being awarded their doctorates in 1976 and 1978 respectively) [5]:-
During the 1960s, Kalman was the leader in the development of a rigorous theory of control systems. Among his many outstanding contributions were the formulation and study of most fundamental state-space notions (including controllability, observability, minimality, realizability from input/output data, matrix Riccati equations, linear-quadratic control, and the separation principle) that are today ubiquitous in control. While some of these concepts were also encountered in other contexts, such as optimal control theory, it was Kalman who recognized the central role that they play in systems analysis. The paradigms formulated by Kalman and the basic results he established have become an intrinsic part of the foundations of control and systems theory and are standard tools in research as well as in every exposition of the area, from undergraduate engineering textbooks to graduate-level mathematics research monographs. During the 1970s Kalman played a major role in the introduction of algebraic and geometric techniques in the study of linear and nonlinear control systems. His work since the 1980s has focused on a system-theoretic approach to the foundations of statistics, econometric modelling, and identification, as a natural complement to his earlier studies of minimality and realizability.
In addition to many research papers, Kalman has jointly authored the book Topics in mathematical system theory (1969) with Peter L Falb and Michael A Arbib. The authors write in the Preface:-
This book does not pretend to be a systematic treatise. Rather, it aims to present a mathematical system theory as it is today - a lively, challenging, exciting, difficult, confused, rewarding, and largely unexplored field. Each of the four sections of the book is independent. No attempt has been made to eliminate duplications; there is no standardized notation.
We mentioned above that Kalman received the Steele Prize in 1986. This was one of many prizes and awards that were made to him as a consequence of his outstanding, innovative contributions. In addition to the Steele Prize he received the Institute of Electrical and Electronics Engineers' Medal of Honor in 1974, the Institute of Electrical and Electronics Engineers' Centennial Medal in 1984, and the Inamori Foundation's Kyoto Prize in High Technology in 1985 as:-
... the creator of modern control and system theory. Kalman theory, which was established in the early 1960s, brought a fundamental reformation to control engineering and since then laid the foundation for the rapid progress of modern control theory.
He also received the Richard E Bellman Control Heritage Award in 1997, and the National Academy of Engineering's Charles Stark Draper Prize in 2008:-
... for the development and dissemination of the optimal digital technique (known as the Kalman Filter) that is pervasively used to control a vast array of consumer, health, commercial, and defense products.
On 7 October 2009 U.S. President Barack Obama honoured Kalman in an awards ceremony at the White House when he presented him with the National Medal of Science, the highest honour the United States can give for scientific achievement. Eduardo Sontag explained the significance of Kalman filtering in [4]:-
Among Kalman's early work was the development of what is now called the Kalman filter for detection of signals in noise. This revolutionized the field of estimation, by providing a recursive approach to the filtering problem. Before the advent of the Kalman filter, most mathematical work was based on Norbert Wiener's ideas, but the 'Wiener filtering' had proved difficult to apply. Kalman's approach, based on the use of state space techniques and a recursive least-squares algorithm, opened up many new theoretical and practical possibilities. The impact of Kalman filtering on all areas of applied mathematics, engineering, and sciences has been tremendous. It is impossible to even begin to enumerate its practical applications. Just as examples of their diversity, one may mention the guidance of the Apollo spacecraft and of commercial airplanes, uses in seismic data processing, nuclear power plant instrumentation, and demographic models, as well as applications in econometrics.
In addition to the above mentioned honours, he has been elected to the National Academy of Sciences of the United States, the American National Academy of Engineering, the American Academy of Arts and Sciences, the Hungarian Academy of Sciences, the Académie des Sciences, and the USSR Academy of Sciences.
Kalman's ideas on statistics are thought-provoking and he has written a number of interesting articles on this topic since retiring. For example he gave an informal lecture published as Probability and science (1993) in which he (quoting from a review by Zeno G Swijtink):-
... laments the explosive growth of applications of probability after World War II, and questions whether probabilities exist in the real world. Randomness does, and to capture it better he proposes a new definition: random is not uniquely determined by simple classical rules. All irrational numbers are said to be random under this definition, as are most "3x + 1" maps. The absence of interaction between chaos theorists and probability theorists also shows, the author concludes, how irrelevant the classical models of probability are to the real world.
In 1995 he published Randomness and probability, and in 2002 a discussion on What Is a Statistical Model? He believes that [3]:-
... the currently accepted notion of a statistical model is not scientific; rather, it is a guess at what might constitute (scientific) reality without the vital element of feedback, that is, without checking the hypothesized, postulated, wished-for, natural-looking (but in fact only guessed) model against that reality. To be blunt, as far as is known today, there is no such thing as a concrete i.i.d. (independent, identically distributed) process, not because this is not desirable, nice, or even beautiful, but because Nature does not seem to be like that. (Historical aside: recall that physicists had thought at one time that aether was such a necessary, unavoidable, appealing, clear and beautiful concept that it must perforce exist; alas, all physicists now living had to learn that such argumentation cannot lead to good science.) As Bertrand Russell put it at the end of his long life devoted to philosophy, "Roughly speaking, what we know is science and what we don't know is philosophy." In the scientific context, but perhaps not in the applied area, I fear statistical modelling today belongs to the realm of philosophy.
Finally note that Kalman married Constantina Stavrou who studied economics at Johns Hopkins University; they have two children, Andrew and Elisabeth.
正文里的方括号编号指向这里,悬停即可直接看到条目。书目保留原文——译了书名反而查不到文献。
原站列出的延伸阅读与外部数据库,照原样保留,目标多为英文页面。
关于鲁道夫·卡尔曼的其它页面:
关于鲁道夫·卡尔曼的其它网站:
原站的交叉引用。指向本站已镜像专题的留在站内,其余仍指回原站。