December 11, 2009
I am greatly honoured to be the recipient of the 2009 CSS Field Award. I have been in our fi eld long enough to have received quite a few honours but I value this award most highly because it comes from our colleagues, specialists in our area, and because previous recipients are such illustrious members of our community; I am honoured to be associated with them.
As I look through the list of recipients, I am reminded of the diversity of their research; this diversity is a great testament to the richness of our field. It was not always so. Charles Desoer, who was an undergraduate in the 1950's as was I, once remarked that when he commenced his research he was disappointed to find, in comparison with physics, so few distinguished texts on control. That is certainly not the case today; our field is now so extensive that relatively few of us have a comprehensive mastery of all of it; our literature is now very extensive. The most signi cant factor in this remarkable development was the revolution of the late 1950's and early 1960's when Kalman's results on fi ltering, optimal control and system theory, Bellman's dynamic programming and Pontryagin's maximum principle burst on the world. It was not only the contributions made by these giants of our field, it was the viewpoint, the way problems were formulated and the way systems were described, that opened up our fi eld so dramatically and liberated us from the strait jacket that then restricted us.
Looking back on my long and enjoyable career, the following experiences come to mind.
The first is luck! Because of apartheid, I left South Africa in 1959, without a Ph.D, for Imperial College London completely unaware of the revolution in control that had already started. The next few years were fi lled with excitement as my horizons were dramatically widened. New problems abounded, fully occupying me and my research students and my research ranged widely. Among the problems we tackled in the early 1960's were optimal control algorithms, recursive Monte Carlo state estimation for nonlinear systems (the forerunner of particle ltering), Monte Carlo methods for optimal control of nonlinear stochastic systems and system identi cation. This exciting encounter with the big bang in control had a lasting influence on me and optimization and optimal control remained major threads in my future research.
A second important factor in my research career was the pleasure and reward of collaborating with excellent research students. I am very proud of having been the Ph.D adviser of this years Bode Prize lecturer, Peter Caines. Another fantastic researcher was David Jacobson, a pioneer inter alia of risk sensitive control, of whom Jason Speyer once said "everything he touches turns to gold" and with whom I co-authored the book "Di erential Dynamic Programming" written after we simultaneously nished our Ph.D theses. I remember opposing very strongly a line David wished to pursue for his Ph.D; he stubbornly ignored my opposition and proved me wrong. Present in the audience is my 40th and last Ph.D student, Sasa Rakovic, who is making important contributions to model predictive control.
A third, and most important, factor is collaboration with colleagues. Research is so individual that generalization is dangerous. I have had the good fortune to work for long periods with exceptional colleagues, Lucien Polak and Graham Goodwin; what made this collaboration work so well was friendship coupled with individual but complementary expertise. Lucien challenged me to acquire necessary analytical skills for our extensive research on optimization and from Graham I acquired a real appreciation for the complexity of adaptive control. But I have also had the good fortune (luck again) to benefi t from the expertise of Martin Clark, Richard Vinter and others in the control group at Imperial College and also from many other colleagues in the world-wide control community especially Jim Rawlings with whom I recently collaborated in writing a book on model predictive control.
Another factor is choice of research topic. This is a personal choice but does depend, at least in part, on the 'dynamics' of our subject. Our subject evolves, I believe, like a Poisson process, with jumps of random magnitude at random times. After the 'big bang' in the late 1950's and early 1960's, there have been many jumps, each off ering a new opportunity for research mainly because of the unexpected widening of our horizon. These jumps arise for di erent reasons. Sometimes the jump occurs when someone is astute enough to observe a problem not adequately addressed by the current theory. For example, after the big bang, we temporarily lost the ability to address the important property of robustness for which the LQR problem was inappropriate. George Zames did not and reminded us of the relevance, in this context, of system gain, addressed through a min-max optimal control problem, and thus triggered the start of H control. A diff erent sort of jump occurs with the introduction of a new framework, such as that in nonlinear control when Roger Brockett pioneered the geometric approach in the early 1970's leading, inter alia, to the comprehensive treatment of nonlinear control in the book by Alberto Isidori. A third type of jump is 'invention' when a new method for dealing with a hitherto intractable problem is provided; examples that are familiar to me are the use, by Goodwin and his colleagues, of normalization to slow down the estimator in adaptive control to ensure convergence and the invention of backstepping, by Petar Kokotovic and his colleagues, for the sequential design of controllers for nonlinear systems. There are other examples. The point I am making is that each of these step changes signals the opportunity for interesting and productive research. Even better would be to make the opportunity by instigating such a change.
The final, and most important factor for me in my career, has been the love and support of my wife Josephine who is with me today. Without her love I could not have achieved the success I have had and I regard this award as much hers as mine.
I am loath to predict the future ... some of my past predictions have been dramatically wrong. Yet, certain developments seem likely, if not certain. New application areas, such as networks, biological systems and control of flocking, will continue to arise and will provide challenges and opportunities to which we, the control community, because of our flexibility and willingness to study new elds, are uniquely quali ed to contribute. Secondly, control theory has become so rich and flexible, that random jumps in our subject will continue to occur, providing us with new insights and new areas to conquer. What Peter Whittle said of research in linear systems applies, I'm sure, to all of control theory: the linear model seems to have in nite depth and yields only to reveal further mysteries ... its know theory becomes more extensive and defi nite with time but, somehow, never de finite.
I thank the IEEE for this award and the control community and the Control System Society for providing such an encouraging research environment.
Thank you all and good night.
