Revision as of 05:12, 10 August 2009
About Petar Kokotovic
Petar Kokotovic is the founder and director of the University of California-Santa Barbara's Center for Control Engineering and Computation. This center has become a role model of cross disciplinary research and education. One of the Center's achievements is a fully integrated cross-disciplinary graduate program for electrical and computer, mechanical and environmental, and chemical engineering fields. While at the University of Illinois at Urbana, he pioneered singular perturbation techniques, used today in power systems and adaptive controllers. At the University of California his group developed constructive nonlinear control methods and applied them, with colleagues from MIT, Caltech and United Technologies, to new jet engine designs. As a long-term industrial consultant, he has contributed to computer controls at Ford and to power system stability at General Electric. With his 30 Ph.D. students and 20 postdoctoral researchers, Dr. Kokotovic has co-authored numerous papers and 10 books. He is an IEEE Fellow and a member of the U.S. National Academy of Engineering. His honors include the IEEE Control Systems Field Award, the IF AC Quazza Medal, the D.C. Drucker Eminent Faculty Award and two Outstanding IEEE Transactions paper awards.
In this account, based on a 1995 oral history, Dr. Kokotovic traces his journey from Belgrade, where he was born in 1934, to the United States. In the process, he pays homage to all of the teachers and mentors along the way that sparked and furthered his interest in cybernetic control systems. As a nonaligned country, Yugoslavia allowed him to study both in the East and West. He first learned about feedback in Paris, gained practical knowledge in industrial technology in Stuttgart, worked with Alexander Feldbaum in Moscow (having left Belgrade in the wake of offending powerful officials), and was finally encouraged by Mac Van Valkenburg to tour the U.S.
Note: Because of Dr. Kokotovic's revision of this oral history into a first-person narrative, it differs in format from other oral histories on the IEEE Global History Network.
About the Interview
PETAR KOKOTOVIC: An Interview Conducted by Frederik Nebeker, Center for the History of Electrical Engineering, 22 February 1995 — Revised for Publication by Petar Kokotovic in February 2008
Interview #246 for the Center for the History of Electrical Engineering, The Institute of Electrical and Electronics Engineers, Inc.
This manuscript is being made available for research purposes only. All literary rights in the manuscript, including the right to publish, are reserved to the IEEE History Center. No part of the manuscript may be quoted for publication without the written permission of the Staff Director of IEEE History Center.
Request for permission to quote for publication should be addressed to the IEEE History Center Oral History Program, 39 Union Street, New Brunswick, NJ 08901-8538 USA, or email email@example.com. It should include identification of the specific passages to be quoted, anticipated use of the passages, and identification of the user.
It is recommended that this oral history be cited as follows:
Petar Kokotovic, "Becoming a Control Engineer 1956-1965," a 2008 revision of an oral history conducted in 1995 by Frederik Nebeker, IEEE History Center, New Brunswick, NJ, USA.
Becoming a Control Engineer 1956-1965
Interviewee: Petar Kokotovic
Interviewer: Frederik Nebeker
Date: February 22, 1995
Place: UC, Santa Barbara, CA
Revised: February 2008 by Petar Kokotovic
What sparked your interest in control engineering: a lecturer, a textbook, a laboratory? What brought you to this career path? All my Ph.D. and post-doctoral students responded thoughtfully, until one of them challenged me: "It's your turn — when and how did you become a control engineer?"
To answer, I look back to my educational odyssey through four European countries, until I landed on the American academic shores. I started as a student seeking industrial practice and continued as a practicing engineer seeking help from theory. Over a period of ten years I acquired a practical-theoretical blend of knowledge that has served me well throughout my career as a researcher, educator and industrial consultant. If my educational odyssey has a message for a young person today, it is to actively seek what, where, and from whom to learn. Furthermore, it is vital to nurture the human bond with mentors and coworkers; research is more fulfilling when pursued with friends than as a solitary endeavor.
To follow this narrative it is useful to keep in mind the horrendous devastation of Yugoslavia in WW2 when more than one tenth of the population perished. My native Belgrade bore the sad distinction of being the most heavily bombed and destroyed Allied city outside the Soviet Union. The four years of Nazi occupation left my generation with practically no elementary education. In the post-war years our high schools lacked teachers, buildings and other basic means. It seemed natural to us that to learn, we had to struggle, which made the hard won knowledge more precious. During my university studies the world was divided by the Iron Curtain and the West was stunned by the Soviet space flights. Many attributed the Soviet success to the superiority of their mathematics and control theory. In addition to my studies in Belgrade, I spent extended periods in Paris, Stuttgart, and, finally, in Moscow. Unexpectedly, one year of study behind the Iron Curtain led me to the United States. In 1965, soon after I received my Soviet degree, I was invited to present a series of lectures at twenty US universities and research institutions. My academic career can be seen as a prolonged continuation of that whirlwind tour.
The Lure of Feedback in Paris
In the spring of 1956, at the age of 22, I was completing my fourth year of the five-year Electrical Engineering program at the University of Belgrade. My interest in electromagnetism evolved toward peaceful uses of nuclear energy, a great promise of that era. A scholarship from the Belgrade Nuclear Institute should have secured my future as a nuclear engineer, but my first real life encounter with feedback directed me along a different path. That this encounter happened in Paris was due to major political changes caused by Yugoslavia's abrupt separation from the Soviet bloc in 1948 and its turn to the West. In the summer of 1956 the students of my generation were given the opportunity to apply for practical training positions at leading companies in Western Europe. My choice was France, whose language and culture I had studied for several years.
In the summer of 1956 I was placed at a power station in a suburb of Paris. Its generators and steam boilers were instrumented with an intricate network of electric, pneumatic and hydraulic sensors, transducers, and actuators, all connected through analog computing elements. Lost in this maze, I asked a friendly instrumentation engineer for guidance and, with his help, discovered feedback and its pervasive role, not only in the regulating loops, but also in many measuring devices. Compensating schemes based on 'force balance principle' were particularly clever. In a pressure gauge, a displacement of its membrane would give a distorted pressure signal due to membrane nonlinearities. However, using a solenoid force feedback to keep the displacement extremely small, almost zero, avoided the distortions and provided an accurate pressure signal. It seemed unreal: a motionless membrane transmits a correct pressure signal in the form of the solenoid current. My excitement with such discoveries competed with other enticements that Paris offered a young foreigner willing to sacrifice hours of sleep.
When I returned to Belgrade, Professor Dusan Mitrovic introduced a course on servomechanisms that covered analysis and design using the Nyquist-Bode and root locus methods. It also included practical projects and laboratory experiments with typical control components. The course marked the beginning of control theory education in Belgrade and I responded like a hungry person served a sumptuous meal. In retrospective, what attracted me to control engineering was the combination of the experience in Paris and Mitrovic's course.
Hard Work in Stuttgart
Early in 1957 I was introduced to a visiting German businessman who soon after called and asked: "Would you like a 'werkstudent' job in Stuttgart, at an AEG factory? AEG is a leading company." After a crash course of German I showed up at the AEG factory where I was given some menial tasks. Searching for more challenging work, I saw a technician testing the regulation loops on an experimental arc furnace. I talked to his boss and ended up testing feedback loops with magnetic amplifiers and vibrators. I knew nothing about these components and, if it were not for the grandfatherly attention of an elderly technician, I would have short-circuited and burned some expensive equipment.
My sleepless nights in Stuttgart differed sharply from those in Paris. It was urgent that I master magnetic and power electronics circuits and learn how to use them to control arc furnaces, electric drives, conveyor belts, and similar industrial systems. The biggest challenge was a vibrating paint spray system for Mercedes cars in a nearby Daimler-Benz factory. Its two inputs and two outputs were highly interactive and the loops were very hard to tune. This brought to my attention the issue of tuning, later to become one of my main interests, including sensitivity tuning, self-tuning and adaptive control.
While testing these nonlinear pulse-width modulated systems I observed that, if tuned, most of them performed well. However, to my knowledge, there was no theory for their design and tuning. For a beginner like myself it was important to witness how a technology developed through experiments, way ahead of theory. Later I would realize that this was not an exception: feedback was used for centuries before Nyquist and Bode finally devised a theory for it.
Belgrade Engineering Diploma
Back in Belgrade I took elective courses in power systems and electric drives. Each included a few control topics such as voltage regulators and motor speed control. The requirement for the 'diploma engineer' degree comprised nine semesters of courses, with the tenth semester dedicated to the completion of a diploma project and its defense. I asked Professor Mitrovic to be the supervisor of my diploma project. He was an inspiring mentor, with inexhaustible creativity and energy. Even as a war prisoner in 1941-1945, he managed to do research 'in his head' and to publish papers via the Swiss Red Cross. His algebraic method appeared in a series of three papers in the January 1959 issue of the AIEE (pre-IEEE) Trans. App. & Ind. and it was included in several textbooks. Mitrovic was also a practicing designer of control systems and analog computers.
After hearing about my work in Germany, Mitrovic told me: "Build your own magnetic amplifier, use it in a positioning servo, and see how far you can analyze and predict its performance." The 'build' part of that task proved to be harder than anticipated because of the low quality of available magnetic material and semiconductors. Luckily, such imperfections accentuated a type of nonlinear behavior which I was able to analyze using the describing function technique I had learned from Control System Synthesis (1955) by Truxal, and Feedback Control Systems (1956, in French) by Gille, Pelegrin and Decaulne. These two textbooks, along with the book Servomechanism Practice (1954) by Ahrend, remained my main sources for several years.
I received my 'diploma engineer' degree in the middle of 1958 and went to work at the Nuclear Institute. My first job was on Reactor 'A,' acquired from the Soviets after Khrushchev normalized relations with Yugoslavia. In my redesign of the servomechanism for tracking the heavy water level, the main difficulty was due to sensor hysteresis, which, to my regret, was impossible to cure with feedback. Before my next assignment on the 'zero power' Reactor 'B,' I was drafted for military service. The draft saved me from participating in the tragic accident when Reactor 'B' went supercritical causing one death and several major injuries. A redeeming feature of the otherwise boring year of military service was the final four-month period when I operated a British made anti-aircraft radar. Playing with the azimuth and elevation servomechanisms allowed me to make comparisons with systems analyzed in the 1947 classic Theory of Servomechanisms by James, Nichols and Phillips. This was a useful bit of hands-on experience with control systems operating in a strict military environment.
Automation Projects in Yugoslavia
After completing the military service in 1959 I was tempted to join the 'Belgrade Hand' project, sponsored by the US Veteran Administration and led by Professor Tomovic, an indefatigable promoter of research. The project's goal was to develop a 'cybernetic prosthesis,' which received wide media coverage fanning the hopes of thousands of WW2 amputees. From my experience with sensors, actuators and other control hardware (decades before VLSI and MEMS) I felt that the publicized promises of the prosthesis project were highly unrealistic. Because of these doubts I turned to other projects whose ambitious goals I considered to be achievable.
Around 1960, the automation laboratory from the Nuclear Institute transferred to the Telecommunications Institute named after Michael Pupin (1858-1935), the inventor of 'Pupin coil' widely used in AT&T networks. From a village near Belgrade Pupin came to New York at the age of 20 and became a professor at Columbia University in 1901. He received the Edison Medal, Pulitzer Prize, advised President Wilson, and was president of IRE, AIEE, and AAAS. In the newly formed Pupin Institute I was assigned to lead the industrial control group of 12-15 engineers and technicians. Situated on a hill overlooking Belgrade, the institute's new building soon became a beehive buzzing with ideas and projects. My group was engaged in projects for food and sugar processing, cement kilns, rolling mills, power stations and railroads. Our tasks included design and implementation of PID control loops, as well as systems with on-off and programmable controllers. A few times we used self-optimizing schemes, inspired by the seminal paper by Draper and Li, (ASME, vol. 160, pp 1-16, 1951) and the 1954 book Engineering Cybernetics by Tsien. In addition to analysis, simulation and design, we also pursued prototype development. The great diversity of our projects helped us gain experience, but it also required hours of hard work.
In one of our projects we were to develop a tuning procedure for the voltage regulators in a large hydro-electric station equipped by Konchar, a major company in Zagreb, Croatia. The leader of the Konchar control group was Joze Cernelc, whom I met in the summer of 1960 and closely collaborated for many years. The regulators had a multi-loop structure, with tunable parameters in each loop. The lack of effective methods for multi-loop structures triggered my first original result. I thought of iterative tuning based on sensitivities with respect to tunable parameters. With a recursive use of Bode's sensitivity formula, I found 'sensitivity points' at which to measure the signals for tuning. This simple result attracted attention and I was invited to visit the research groups of W. Oppelt in Darmstadt, E. Gerecke in Zurich, P. Eykhoff in Eindhoven, and J. Westcott in London. While designing and testing controllers for multi-loop systems I faced the danger of unmodeled dynamics that caused high frequency oscillations in inner loops and threatened to damage expensive rotating amplifiers. Several years later I learned how to parametrize various types of unmodeled dynamics as singular perturbations. This became a research topic that was to occupy a significant part of my career.
Other versions of sensitivity tuning, which my group developed for more complex systems with nonlinearities, were also successful in practice. For a General Electric equipped rolling mill in Sisak, Croatia, our tuning reduced the reversal time and increased the mill's output. This result brought me in contact with Harold Chestnut, a leading control engineer at General Electric and the founding president of IFAC. Our automation projects for the iron and steel industry continued for several years and taught us that socio-economic issues of automation are often more complex than our technical problems.
Post-Diploma Studies in Belgrade
The late 1950's and the early 1960's saw the extremely rapid growth of 'modern control theory,' which included state space models, Lyapunov stability and the optimal control twins: Maximum Principle and Dynamic Programming. Concepts of self-optimizing and adaptive control were also gaining ground. All of them shined at the First IFAC Congress in Moscow, 1960, where I feverishly rushed from one inspiring session to another. I attended this historical event by chance, as a substitute for my boss, due to his sudden illness. The scientific grandeur of the Congress was almost eclipsed by the lavish banquet at the Kremlin given by Khrushchev and his team.
My colleagues at the Pupin Institute and I were determined to follow these exciting theoretical developments, in addition to our full time work. The university of Belgrade was about to start post-diploma studies leading to the 'magistar degree,' somewhat more advanced than the US master degree. We submitted a list of subjects which were included in the program of evening courses covering functional analysis, stochastic processes, ODE's, PDE's, calculus of variations and similar mathematical subjects. The newest control topics were left for self-study, seminars and the summer school held in Dubrovnik, a historic town on the Adriatic Coast. Among the lecturers were Richard Bellman, Lotfi Zadeh, Revaz Gamkrelidze and other leading researchers.
My upbeat mood of that period came to an abrupt end by the tragic death of Mitrovic, at the age of 50. Without him my 'magistar' thesis defense was filled with sadness. Mitrovic influenced my views on many aspects of science, society, and politics. His wit was frequently directed at corrupt bureaucrats ('comrades') who were common and powerful in Yugoslavia. There was more than an echo of Mitrovic's sarcastic tone when I asked powerful comrades at an 'anti-corruption' meeting about their personal use of public property. They reacted with visceral anger and ominous threats. Soon after, a friend warned me about pending repercussions and urged me to find a pretext to leave the country for an extended period. "You don't have time to wait for Fulbright. Get a Soviet fellowship, they have more openings than applicants, only one violinist and two ballerinas applied." I was no longer single and it was very hard to leave my family behind.
The Year in Moscow
The two-day train ride to Moscow gave me the opportunity to think how best to use the unexpected year of study abroad. My desire was to work with Alexander Feldbaum, a pioneer of optimal control and self-optimizing systems, well known in the West for his 'dual control' theory. But, I was to face a huge roadblock: Feldbaum worked at the famous IAT, Institute of Automation and Remote Control, which was strictly closed to foreigners.
Upon my arrival in Moscow in February 1964 I was placed at a teaching institution with no research activity in my area of interest. I felt lost and purposeless; all my calls to Feldbaum's office at IAT were brusquely answered: "He is out of town." And then, almost by accident, I spotted an announcement for a public lecture by Feldbaum. After the lecture he was friendly and direct: "You know, I haven't been out of town. I read one of your papers and would like to discuss it with you. Come tomorrow, I will meet you at the main entrance." Each morning Feldbaum would escort me through the IAT's elaborate security, until one morning I dared to walk in on my own. It worked! If curiosity worked in Paris and perseverance in Stuttgart, Moscow called for audacity! Ultimately the IAT security issued me a retroactive permit.
IAT was the oldest and best known control theory institute in the world and Feldbaum was one of its leading researchers. He seemed to know everything: from stochastic decision theory to technological details of components. "I am an engineer, but I also have a degree in math just to help me communicate with mathematicians." And communicate he did, most notably when he formulated the optimal control problem for Pontryagin and coworkers, who soon after produced the celebrated Maximum Principle. Tragically, at the very start of my studies with Feldbaum, he suffered a devastating stroke which incapacitated him until he died in 1969, at the age of 56. My shock was softened by Feldbaum's friend Yakov Tsypkin, who took me into his group.
Tsypkin had received the highest Soviet recognitions already in the 1950's. By the 1960's he was internationally known through his papers and books on relay and sampled-data systems. His pioneering results were often cited as 'Tsypkin's method,' 'Tsypkin's criterion,' etc. Although I studied under his guidance for only one year, he continued to nurture me with his optimism and friendship until his death in 1998. We would regularly meet at international conferences and he visited me several times in Illinois and California. The last time we spent a few days together was in San Francisco during the 1996 IFAC congress.
It was thanks to Tsypkin that during the year I spent at IAT, from February 1964 through April 1965, I was able to receive my 'candidate' degree, the Soviet counterpart to PhD. "You already have several papers published in good journals. If you extend your existing results, give them a clear focus, and also pass our exams, you can get a candidate degree at IAT, as an exception to our rules." To hear this from Tsypkin was encouraging, but then he continued: "A lot is going on in our field these days. At seminars you can gain breadth and get ideas about new research topics. Work hard on both, your thesis and breadth!" Tsypkin's own breadth was immense. With an encyclopedic grasp, he would examine every new result to find its roots in a fundamental concept. For example, talking about Popov's 1959 breakthrough on the problem Lurie posed in the 1940's, Tsypkin revealed, in a few transparent steps, the passivity basis of Popov's stability criterion. Then he informed me: "Lurie will be on your committee." Presenting results to Anatoly Lurie was an awe-inspiring experience. The work of the 'Leningrad school,' founded by Anatoly Lurie and subsequently led by Vladimir Yakubovich, set the stage for many significant developments in the following decades, including the LMI design methods widely used today.
My 'breadth gaining' studies, some of which were required, encompassed optimal control, nonlinear stability theory, pattern recognition, information theory, etc. Tsypkin helped me to organize this multitude of new concepts and theories. I also received generous scientific and moral support from Boris Kogan, director of IAT's analog and hybrid computation laboratory. He is now associated with UCLA and, in his 90's, passionately continues his research activities. Lively conversations at IAT lunches were often fascinating. At one of them, Mark Aizerman, of the 'Aizerman conjecture' fame, talked about an algorithm in his laboratory that distinguished between Mozart and Bach. Tsypkin asked: "Are perceptron-like algorithms special cases of a general least squares method?" Two years later that was the theme of Tsypkin's plenary talk at the 1966 IFAC congress in London. At lunches I heard about the groundbreaking work of Nikolai Krasovskii and his school at Sverdlovsk. When at a lunch I asked which ballets, concerts and theater performances to attend, Tsypkin teased me: "Is your 'breadth' getting too broad?" Alexander Letov, who was fluent in English and widely traveled as IFAC President 1957-60, advised: "Skip Moscow's boring theaters, wait until you get to Broadway."
Tsypkin had interest in sensitivity analysis and saw my tuning ideas as fitting in the stochastic approximation framework he used for his theory of 'learning systems'. Motivated by the unmodeled dynamics I encountered in practice, I wanted to expand my thesis to include 'the sensitivity to the change in system order,' that is 'singular perturbations'. Tsypkin thought this to be a good program for my future research, but advised that I not try to include it in my thesis. What I learned during that period from the works by mathematicians Tikhonov and Vasileva I used later in my singular perturbation research.
Among the requirements for Soviet degrees set by the Ministry of Higher Education was a Marxism exam. This requirement was removed for other non-allied countries, like India, but as a Yugoslav I had to petition to the Ministry. When I heard that the petition process could take months, I decided to take the exam, which led to an amusing situation, characteristic for that period. The exam was oral with questions typed on the cards in a box in front of me. By a strange coincidence, the card I chose at random was about 'the Yugoslav revisionist Marxism'. My answer was immediate: "In 15-20 minutes I can prove that, on the contrary, the Soviet version of Marxism is revisionist. Should I continue?" My examiners preferred not to accept this challenge: "No need for further discussion, you pass!" At my thesis defense, the secret vote of the IAT scientific council was overwhelmingly positive, and after a Russian-style farewell party, I was on my way to Belgrade by train. When the train was about to exit from the USSR, the Soviet border officials questioned the validity of my visa and ordered me to get off. With a final bit of audacity I refused and arrived in Belgrade by the same train. There I found, to my great relief, that the comrades had forgotten me.
Final Destination: USA
If this brief account of my arrival to the United States reads as homage to Mac Van Valkenburg, or simply Mac, it is because he initiated my American career and served as my academic mentor for many years. Mac was a professor, and then Dean of Engineering, at the University of Illinois. He was a popular leader of engineering education and published several widely used textbooks. His Network Analysis, published in 1955, broke all publishing records. We met in the summer of 1964 at the IFAC Sensitivity Analysis symposium in Dubrovnik, which I attended with seven IAT members led by Tsypkin. These Soviet 'Magnificent Seven,' as they jokingly called themselves, cordially interacted with their American counterparts.
Tsypkin and Van Valkenburg included me in some of their conversations, and I enjoyed observing their similarities. Both liked to inform and explain, sometimes to entertain, but never to impress. As mentors, both considered it natural to unselfishly spend hours and days helping their younger colleagues. It was my great fortune that my future was charted by these two remarkably generous men, residing on the opposite sides of the Iron Curtain, but cheerfully ignoring its ugly existence.
During the Dubrovnik symposium Mac offered to organize a lecture tour in America for me after I completed my studies in Moscow. In a letter I received from him when I returned to Belgrade in the early summer of 1965 he proposed a three-month tour of about twenty leading universities and several research centers. When I expressed my concern as to whether what I had to say would be of interest, 'Mac the Mentor' wrote to dispel my fear and encourage me: "Your lecture topics are excellent and speakers with first-hand knowledge of Soviet research are extremely rare." As if to encourage me a bit further, the ambassador of US in Belgrade invited me to join a small group of guests at his dinner table. Finally, with an AFOSR prepaid ticket I boarded a PANAM flight to New York City, where my tour started. At the tour, the post-Sputnik era curiosity and the topics of current interest worked in my favor. I was warmly welcomed by large audiences and engaged in lively technical discussions.
The tour was an intense learning experience for me, scientific and cultural. It brought me in contact with colleagues with whom I would collaborate, or keep in touch, throughout my career. In Urbana, Illinois, where I stayed for two weeks during the tour, and then for another 25 years, the contacts with Mac Van Valkenburg, Joe Cruz, Bill Perkins and Dale Compton have since grown into lifelong collaborations and friendships. At Santa Clara University in California I celebrated the reunion with Drago Siljak, my former classmate and close friend from Belgrade.
During the tour, discussions with numerous colleagues opened new research vistas for me: Peter Dorato, Rudi Drenick, and Dante Youla at Brooklyn; Bob Narendra at Yale; Mike Athans, Roger Brockett and Hans Witsenhausen at MIT; Larry Ho at Harvard; Mike Mesarovic and Sanjoy Mitter at Case; Steve Kahne at Minnesotta; Jim Massey and Mike Sain at Notre Dame; Dick Bucy at Colorado; Boyd Pearson and Angelo Miele at Rice; Lotfi Zadeh, Charlie Desoer and Lucien Polak at Berkeley; Rudi Kalman, Tom Kailath, Gene Franklin, Bernie Widrow, Bob Newcomb and Brian Anderson (student) at Stanford; Richard Bellman and George Bekey at USC; Harold Chestnut at GE; Irv Sandberg at Bell Labs; Jorma Rissanen at IBM; among many others.
The impact these colleagues had on my future career was decisive. I felt swept into the flow of their ideas, concepts and theories. Decades later, I am still part of that flow.