Oral-History:A. Stephen Morse

About Stephen Morse

A. Stephen Morse was born in Mt. Vernon, New York. He received a bachelor's degree in Electrical Engineering from Cornell University in 1962, a master's degree from the University of Arizona in 1964, and a Ph.D. degree from Purdue University in 1967. From 1967 to 1970, he served a two year term of active duty (plus an additional year) with the Office of Control Theory (OCTA) at the NASA Electronics Research Center. After leaving the NASA lab in 1970, he came to Yale University where he is presently the Dudley Professor of Engineering. Morse served as a visiting scientist at the University of Toronto, as well as took part in several sabbaticals and leaves. including at Caltech, Australia, Osaka, and Florence.

Morse's research interests include control theory and adaptive control, systems theory, sensor networks, and control of mobile autonomous agents. For his work he has received several awards and honors, including George S. Axelby Outstanding Paper Award in 1993 and 2005, American Automatic Control Council's Best Paper Award, and the 1999 IEEE Technical Field Award for Control Systems.

In this interview, Morse discusses his educational experience in engineering and his gradual introduction to, and involvement with, control theory. Outlining his appointments at the NASA research lab and at Yale, he reflects on his research, and the challenges and evolution of the field throughout his career. Additionally, he points out his various collaborations and influences, provides advice to young people interested in the field, and muses on the future of control theory.

About the Interview

A. STEPHEN MORSE: An Interview Conducted by Brian Anderson, IEEE History Center, 22 June 2015.

Interview #753 for IEEE History Center, The Institute of Electrical and Electronics Engineers Inc.

Copyright Statement

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 Director of IEEE History Center.

Request for permission to quote for publication should be addressed to Oral History Program, IEEE History Center at Stephens Institute of Technology, Samuel C. Williams Library, 3rd Floor, Castle Point on Hudson, Hoboken NJ 07030 USA or ieee-history@ieee.org. It should include identification of the specific passages to be quoted, anticipated use of the passages, and identification of the user.

A. Stephen Morse, an oral history conducted in 2015 by Brian Anderson, IEEE History Center, Hoboken, NJ, USA.

Interview

INTERVIEWEE: Stephen Morse
INTERVIEWER: Brian Anderson
DATE: 22 June 2015
PLACE: Yale University

Video

Early Life and Education

Anderson:

My name is Brian Anderson and I’ve got a most pleasant duty today, which is to interview one of the leading members of the control engineering community. Rather than name him, I’m going to ask him to introduce himself, but I will say it’s June the 22nd, 2015 and I am at Yale University. And now I will ask the interviewee to please introduce himself and tell us what his position is.

Morse:

Hi. My name is Stephen Morse. I’m a professor of electrical engineering at Yale University.

Anderson:

Thank you very much, Steve. And it’s been a long route coming here and it probably started back in your school days. Would you like to tell us something about your time as a schoolboy even and what led you into this long journey to where you are today?

Morse:

Sure. I think at the time that I was in high school, engineering was very high on the totem pole. It was a very respected subject throughout the world, especially in the United States, and it was attractive to me. I was always kind of a technical person and so I decided to go into engineering, although I wasn’t completely sure what engineering was I must say. When I applied to Cornell, one of the things that you had to do was to declare which field of engineering you would go into. I recall that I didn’t really know what an electrical engineer was or what a chemical engineer was. I was torn between the two, not knowing what either was. I flipped a coin and electrical engineering won. So I matriculated at Cornell as an electrical engineer.

Anderson:

Well, we’re very lucky. Can I assume that there were decisive people in your family or teachers at your school who helped steer you in this direction?

Morse:

There was absolutely no one.

Anderson:

That’s interesting. Why did you go to Cornell?

Morse:

Because Cornell was and is a top engineering school in the United States, so that’s why I chose it.

Anderson:

Was it a big surprise when you actually got into the course? Was it very different from what you anticipated?

Morse:

Well, as I said before, I didn’t know really what engineering was all about. I was overwhelmed. It was a very rough-and-tumble course. I should say that the year that I entered Cornell, the Russians launched Sputnik. This affected everything. American engineering education was under a lot of fire because the United States was not in space first. So all engineering schools across the country were really tightening the screws. There was criticism of engineering being too practice oriented, not scientifically oriented enough. Curricula were changing. As a matter of fact, the curriculum that I was exposed to was changing right under my nose. I started out with courses in rotating machinery and welding and the strength of materials and even drafting. I carried a slide rule. By the time that I left Cornell, after a five year program, everything had changed. I went back to Ithaca maybe ten years after I graduated to give a talk and I decided to bring with me the engineering catalog which had in it the curriculum I encountered when I entered Cornell. I must say it was a big hit as I read some of the courses that had been offered at the time that I entered. I took 16 credit hours of rotating machinery, for example.

Anderson:

So that would be very unusual for a student today, wouldn’t it?

Morse:

It certainly would, yes.

Anderson:

And compared with what the courses were like ten years after you graduated, would it be very different today for an undergraduate?

Morse:

Well, as I said, engineering education at least at Cornell, I think most every place, became more science-based and, in fact, the faculty at most schools changed from people who had served some time in industry to people who had gone through PhD programs. So everything changed. Yes, probably the details of courses changed, but generically I think things were pretty much the same.

Anderson:

Right. And then sometime during your experience at Cornell, you took the decision to do graduate work.

Morse:

That’s correct.

Anderson:

And this is a decision that many undergraduate engineers take one way or the other and it’s a major decision. How did you take that decision and why did you go where you went?

Morse:

To tell you the truth, I don’t remember why I made the decision. I had been brought up in a family where the emphasis was on making money. I think when I thought of graduate school, I thought I would get a master’s degree and then go out and make money, so maybe that I felt that I needed to get a master’s degree. I went for a master’s at the University of Arizona in Tucson because while I was at Cornell, I had been exposed to a course in network synthesis taught by Nick Declaris out of a book by Ernst Gullemin on network synthesis. That’s an absolutely beautiful subject and it’s sad that it is not being taught anymore, but it’s a really totally beautiful subject. I know you know a good deal about that yourself. I was smitten by it. It was one of the first courses that I took that I really was comfortable with because it was more analytical than a lot of the others. So I wanted to go on with that. I knew that at Arizona there was a fellow named Larry Huelsman, who was pushing a subject called active network synthesis. I wanted to have exposure to it. So that’s why I went to Arizona. Arizona was also the place where Motorola had labs and so there was a connection between the University of Arizona ...

Anderson:

Oh, I understand.

Morse:

... and Motorola. Lots of things were happening. It was also a time when electrical engineering was moving away from vacuum tubes and voltage-regulated power supplies to transistors and in particular to integrated circuits. So the idea of building active, network devices made more sense. You wouldn’t want to do that with vacuum tubes. They were too big and too unreliable.

Anderson:

And, and most active network synthesis tried to do away with inductors, didn’t it, which was another kind of embarrassment from the early ages of electrical engineering when we moved into the integrated circuit.

Morse:

That is correct. As a matter of fact, when I was in Arizona, two things happened. I took courses from Huelsman. It was my first real exposure in linear algebra, which I enjoyed, but I also took a course from a guy named Granino Korn. He was an analog computer guy and very good at it. It was my first exposure to operational amplifiers, which up till that time were being used primarily as analog computer devices. They weren’t being used as circuits. So I had the idea along with a number of other people: Hey, why not try to use these things as circuit components, not just as part of an analog computer, especially since they were now being fabricated as solid-state devices and they were small and more reliable? I was very impressed at how easy it was to build a transfer function out of operational amplifiers and how painful it was to do the same thing using a passive network. So that’s what got me going on this. I ended up doing a master’s thesis on the use of operational amplifiers in active circuit theory and in particular I came up with a circuit for a gyrator. A gyrator is a device which you can use to avoid having to use an inductor. I think the other thing I developed was a circuit using operational amplifiers for a negative immittance converter, but the gyrator was the main contribution of my master’s work.

Anderson:

Yes, well, that’s interesting. And, it has some connection to control, but something must have sparked your decision to change universities and change fields, even if it was to an adjacent field, so how did that come about?

Morse:

Well, that something was a guy named John Gibson. John Gibson, also known as Jack Gibson, was a professor at Purdue who came down to Tucson to give a lecture. He basically recruited me. I don’t remember exactly how he did it, but he did do it. I have to say that when I first started in control, even though control and circuits are very closely related fields which have a lot of concepts that are the same, there was something very weird about control, at least there was when I first started learning about it. Now I take it for granted now. That weird thing is a plant. What is a plant? Why should one have a piece of a system that sits there and you’re not allowed to touch it? You know, if you’re building an electrical circuit, you can put your hands all over the circuit and change anything. But in a core control system, there is this subsystem called a plant which is given to you and you can’t modify it. It did take me a while to get used to that idea and I think other people as well initially have trouble grasping the concept.

Anderson:

Yes. So you really moved to Purdue because you were inspired by the message that Jack Gibson gave.

Morse:

Yes, that’s right.

Anderson:

But you didn’t work for him at Purdue, did you?

Morse:

I did, yes. I started out with Gibson. There was a very strong group when I went to Purdue. Some of the names you probably will know. Boyd Pearson was a young assistant professor who taught me a full course on linear algebra out of Gantmacher’s Matrix theory and I think he was having as much difficulty as I was trying to understand all of that subject, especially elementary divisors. In the math department was Len Berkovitz, doing optimization theory. Zeno Rekasius was there. K. S. Fu was there and a fellow named Sid Sridhar, who unfortunately passed away at a young age; he was a very powerful researcher and he moved on to Caltech while I was at Purdue. And Violet Haas, who was also there.

Anderson:

Yes.

Morse:

So Gibson was my professor initially. Then in the middle of my stay there he left for another university and Violet Haas took over and I finished with her.

Anderson:

Yes. Okay, I think you’re known for having had her as the supervisor.

Morse:

Yes.

Active Duty at NASA

Anderson:

Oh, yes. I hadn’t heard about Jack Gibson. Well, when you graduated from Purdue, you had a decision to make where to work. So how did all of that play out?

Morse:

Well, I had half a decision. I had been in ROTC, as many people of that era were. I think by the time I got my PhD, I was a first lieutenant. The Vietnam War raging. So to some extent I didn’t have a choice as I was obliged to go on active duty. But it turned out that Bill Wolovich and a fellow named George Kovatch came to visit Purdue from a place called the NASA Electronics Research Center in Cambridge, Mass. They were looking to hire people. I told them my situation, that I had to go on active duty, and they told me that sometimes people with PhDs can be assigned to a government lab. Indeed, that’s exactly what happened to me. So I spent my two years on active duty at a NASA lab, namely the Electronics Research Center in Cambridge, Massachusetts, which was fortuitous.

Anderson:

Yes. It sounds as if from a very young age you took some very good decisions in terms of the ultimate outcome from those decisions. And some people would call that luck. Others might say that you have luck because you’re well-prepared for it or you can seize initiatives. How do you put yourself in the spectrum looking back?

Morse:

I think I was a very lucky camper. I was lucky to be able to go to the NASA center. This center was a real hotbed of activity. When I showed up there, George Zames was there on a Guggenheim fellowship. Murray Wonham was there on another fellowship. He was on leave from Brown University. Bill Wolovich was there. Other people who would show up there regularly included Peter Falb, Harold Kushner, Mike Athens and Roger Brockett, so it was a very, very exciting, active place. Not long after I got there, George Kovatch and Hugo Schuck formed a new group called the Office of Control Theory and Application or OCTA. All of the people that I just mentioned were either part of it or affiliated with it. Many distinguished people from outside showed up, including Lotfi Zadeh and Rudy Kalman....and you! So it was the place to be. In terms of luck for a young, a young person like myself, it was just wonderful to have exposure to all of these people and all of these activities at the same time.

Anderson:

It’s a very important principle, isn’t it, to get exposure to some-

Morse:

Oh, boy, gosh, yes.

Starting at Yale University

Anderson:

-the leasers in the profession when you’re at a young age. So your appointment at the laboratory finished at the end of the planned time. And it was at that time that you moved to Yale.

Morse:

Well, yes, it was, but it wasn’t so much that my appointment finished. I was at the NASA center from 1967 to 1970. The first two years I was in the military. Then I was discharged honorably. The third year I decided to remain at NASA. It was a pretty good place to be. I was very happy there. Unfortunately, for various political reasons - and I don’t mean political reasons within the control field - I mean national political reasons - a decision was made to close the center. So all of the technical people, myself included, had to look for jobs. It was more or less like we were pushed out of the center. It was at that time that I went job hunting and was fortunate enough to get an assistant professorship at Yale University.

Anderson:

And you’ve been here now about 45 years.

Morse:

Yes, I guess.

Anderson:

So during that time, you haven’t stayed at Yale 12 months of the year. You’ve had leaves and sabbaticals and visits of a month’s duration. Would you like to comment on some of those and how important they’ve been in your career?

Morse:

Well, I’ve thought about them. I must say, I’ve had to think hard because one tends to forget a lot of these things. I don’t travel as much as you do, but I have taken leave. Yale is very generous in that they give what’s called a tri-annual leave. Every sixth semester one gets a semester off at full pay. It’s a very nice policy. I’ve always taken advantage of it, but I haven’t always gone away from Yale. But when I do, it’s generally not for the whole semester. It would be for a three-month period. I just don’t like to be away for a long period. But I have been to places like Caltech to teach courses. I have spent time in Osaka, Japan as a Japan Society for the Promotion of Science fellow. It was a very exciting experience to be with Hidenori Kimura for a while. During that particular tri-annual leave I was also at the University of Rome for several months with Alberto Isidori. These were very happy times for me. I learned a lot from these people and there was some collaborations in some instances. I’d have to look at a piece of paper to remember more of the places that I’ve spent time such as Florence, Italy with Edoardo Mosca, Jan Willems and Jan Willem Poldermann in the Netherlands, and many other places in Europe. I’ve also spent time in many locations in China and of course Australia with you.

Anderson:

And by the same token, you’ve been host to many people too.

Morse:

I should say Australia more than anyplace. I generally go for a two or three week period. Canberra as you know is a wonderful place to work. It’s peaceful and quiet and nice accommodations and it’s just a very, very nice, comfortable place to work.

Work in Adaptive Control

Anderson:

I believe you were introduced to a practical form of adaptive control on the Canberra Lake? Do you recall what I’m talking about?

Morse:

Yes, I do. You’re talking about windsurfing.

Anderson:

Yes.

Morse:

Yes, I do remember that, yes.

Anderson:

Because adaptive control was one of the major areas of your interest over a number of years, wasn’t it?

Morse:

Yes, that’s correct, about ten years, yes.

Anderson:

You acquired a very, very strong reputation for your work in that area. What led you into that area?

Morse:

I think just being anywhere in the vicinity of Bob Narendra, you can’t help but wonder what’s going on in adaptive control. Bob had been doing adaptive control for many, many years before I came to Yale. So I became intrigued by it also. Prior to coming to Yale I had some exposure to adaptive control at Purdue with Jack Gibson and at the NASA center with Alan Pearson during his visits from Brown University. It intrigued me. Dick Monopoli, another figure in adaptive control, was close by at the University of Massachusetts, so I couldn’t help but get interested in the subject. This ultimately led to an enormously exciting period in the late 1970s when there were players all over the world who were all trying to solve the same adaptive control problem,. In Sweden it was it was Karl Astrom and his student, Bo Egardt and in Australia it was Graham Goodwin and his student Peter Ramadge, while at Yale both Bob and I were trying to find solutions. It was a very exciting period in the late 1970s.

Anderson:

That’s a fair while in the past, isn’t it? And from time to time you’ve probably reflected on the question did adaptive control deliver everything that people wanted from it and is it now a dead area from the point of view of research? What’s your view about the current state of play in adaptive control?

Morse:

Well, I would certainly not say that it’s a dead area. What I would say is that at the present it is not being researched as much as it had been in the past. That doesn’t mean that there aren’t major issues that need to be resolved. I happen to think that adaptive control from an engineering perspective has been one of the most innovative parts of all of the control field to have emerged since the Second World War. There were very, very clever ideas introduced, lots of them by lots of very bright people, but they’re engineering-oriented and less mathematical. Maybe they’re a little bit less appealing to some of the more mathematically-inclined people in the field. So adaptive control has evolved to a certain point, but there remains several significant issues. The first one is that there is yet to be a clear and crisp description of exactly what a parameter adaptive control system is, which can be easily understood by people who are not in the adaptive control community. But even more important, the number one thing that’s missing in adaptive control is a performance theory. There is no credible performance theory at present and those who have tried to build a real adaptive control system know how painful it is to have to do it by trial and error. Actually I had a student whose name is Ming Cao, now at the University of Groningen, who did a real adaptive control system design at Yale and then went down at the GRASP Lab at the University of Pennsylvania to implement it on a team of mobile robots. If you ask him to take off his shirt, you’ll see all of the scars on his body from all of the pain that he had to go through to get things working. So what’s missing is a credible performance theory. Some bright person will come along and refuse to accept conventional wisdom and then will come up with a new idea and there will be a lot of excitement again.

Anderson:

Yes. But then I’m struck by the fact that if you go to a CDC, these days there will be fewer sessions on adaptive control and than there used to be.

Morse:

Right.

Anderson:

And I suppose people thinks it’s a very hard area in terms of the problems that are still left on the table.

Morse:

Well, like most areas, it’s in need of a good new idea, an out-of-the-box idea, which has not emerged yet.

Early Research Projects

Anderson:

Yes, okay. Before you did adaptive control, you did some other things.

Morse:

That’s correct.

Anderson:

What were those?

Morse:

Yes, I want to say several things about that. When I was in graduate school at Arizona I did active network synthesis. When I got to Purdue it was the beginning of the space program and there was a lot of interest in the effects of disturbances on control systems. What I did was to look into designing control systems optimally, from a worst case point of view. I got involved with things like differential games and stuff like that. But the really important message that I want to convey about that time was that this was an era in the 1960s when there was I would say, an unbelievable obsession with optimization theory. If you couldn’t solve a control problem “optimally” you were nowhere. By the time I finished graduate school, I was turned off by the subject and was wondering if there was any more in life after optimal control. Then I went to the NASA center where I immediately encountered to Murray Wonham, who had just finished his pole placement theorem, which had absolutely nothing to do with optimal control. I was singularly impressed with that, not just the theorem itself, but that the pole-placement problem was fundamentally different than an optimization problem. In truth from that day to this, I personally have not worked on optimal control problems. I did get involved with Murray and it was one of the more exciting periods in my life. He had started to work on something called a controllability subspace as a result of having sought to find out when one you do pole placement on a subspace of the state space. I was interested. I entered a dialog with Murray with my background in dealing with disturbances. There is something called the “principle of invariance” which had been devised in the Soviet Union. It is concerned with how you might design a control system to reject externaldisturbances. So Murray and I started to talk and what emerged was what is now known as the “disturbance rejection problem.” Back then we called it disturbance isolation problem. At first, like with most research undertakings, we had no idea what we were doing. Eventually it became clear that there was a very close connection between that problem and controllability subspaces. In fact, the connection is something called an (A,B)- invariant subspace or a “controlled invariant subspace.” That was an enormously exciting period because things happened very, very fast. We worked through that problem developing the basic building blocks of linear geometric theory, but the game wasn’t over because down the hall at NASA were Bill Wolovich and Peter Falb. They were hammering out final solution to their non-interacting control work or the decoupling research. We began to realize that we could also do that kind of thing in the linear geometric theory we were developing. So we eventually formulated and dealt with quite a large class of decoupling problems. All of that was a very satisfying and nice experience. It was after that experience that I went to Yale and then got into the adaptive control business for a while. Then I guess I’m trying to remember when the next change maybe came. I hope I’m not missing something, but the next significant change didn’t come till the beginning of the new millennium. What had happened was the National Science Foundation had started to encourage large multi-disciplinary projects. There was a program called Knowledge and Distributed Intelligence or KDI.

Anderson:

Some buzzwords.

Morse:

Yes, some buzzwords. So I was thinking that it would be nice to write a big proposal in this area. It just so happened that I was attending an ONR research meeting in Monterey Bay, California, where people were talking about swarming and things like that. I had doubts in my mind. I had a layover, a one-night layover. I went to the Monterey Bay Aquarium. There I stood bedazzled in front of a huge fish tank watching schools of what I think were sardines bifurcate as predators would dive into the middle of them. I got an idea. Why not write a KDI proposal on how fish school, birds flock, etc. and how we might use the traffic rules discovered to coordinate the motion of large groups of mobile autonomous agents such as robots? I recruited a team to write such a proposal. The team consisted of Roger Brockett, Naomi Leonard who had already been studying underwater gliders, Peter Belhumeur who did computer vision, and two individuals from the biology community at the University of Washington, namely Julia Parrish and Danny Grunbaum, both experts on fish schooling. This was a very, very strong team and we were fortunate enough to get funded. I think lots of very good research came out of that project. It was during that project that my student Jie Lin and my post doctoral fellow Ali Jadbabai started to work on flocking problems and co-authored one of the first papers on the subject. This was a very exciting experience. It was really a new thing since at the time people were not using graphs very much in control problems.

Anderson:

And at the CDC meeting in 2000 there were probably almost no sessions on multi-agent control ...

Morse:

That’s correct.

Anderson:

... or distributed control ...

Morse:

That’s correct.

Anderson:

.... or control over networked systems...

Morse:

Yes.

Anderson:

And people who have just been round in the last five years would find themselves almost in a different universe if they were to go back to that era.

Morse:

That’s correct. And so I should say that prior to 2000, prior to this period, if you were to look at what was a control problem, it would be consider the differential equation and then do something to it. But generally, it would be a differential equation. But the idea that there would be some structure or some graphs or networks really was not in the mainstream, although there had been work that went all the way back to the 1970s, but most people in control just didn’t think of a control problem as having a graph as an ingredient of the problem. That changed due to a number of people. This continues today and is even increasing as a matter of fact. It’s very exciting.

Anderson:

And do you think that the funding agencies were subsequently responsive to what you produced from that very first grant? Did it have a decisive effect on NSF or AFOSR or other agencies?

Morse:

I don’t know what was going through the minds of these people, but there have certainly been a large number of projects, which have involved multi-agent systems and that continues today. I can think of one project called Swarming that I was involved with at the University of Pennsylvania. That was an ARO MURI so I think, yes. But it wasn’t just our work at Yale that caused the change. There were plenty of other people. Richard Murray and his post-docs were also contributing early on. Naomi Leonard was heavily committed and other people as well. And of course you were too. The field of distributed multi-agent systems really exploded in a very, very nice way.

Students

Anderson:

You’ve mentioned quite a number of people who are famous in our community and you’ve had the good fortune and the skill and your own level of expertise to interact very successfully them. You’ve also had a number of students. Can I switch to the students and say can you highlight some of the things your students have done or name some students who would be well-known in the community today?

Morse:

Well, let me start with a guy I already mentioned, namely Jie Lin {a.k.a. Archer} whose research led to the flocking problem. Keep in mind that at the time he was my student there was almost nothing at all going on with multi-agent systems - it was an pretty empty playing field you might say.

Anderson:

Yes. What happened in the research, which led to this flocking paper was that you wrote with him?

Morse:

He was physicist. He knew that he had to come up with a thesis topic and was very good at searching around the internet. In Physical Review Letters he ran into this article, written by Tamas Vicsek and several other Hungarians, which captured all of our imaginations. We set the problem up as a linear problem. It looked like a discrete time switch linear system, so it was a time varying system. We started to try to analyze the system, but you know that time varying systems ain’t for sissies {to use Betty Davis’s famous expression}. They’re tough to analyze when it comes to stability and convergence questions. We were really stuck for a while. I urged Ali Jadbabai and Archer to drop the problem it because I didn’t want them to waste their time on something which seemed to be so difficult. Well, Ali’s a stubborn guy and he wouldn’t drop it. Instead he hung in there. Eventually he was able to find a theorem of Jacob Wolfowitz, and a fellow named M. Artzrouni in France who was able to help him use it. With that theorem in hand, we were able to finally address the problem directly. So between the post-doc and the graduate student, things went very well and I’m very pleased. Now prior to that, Daniel Liberzon and Joao Hespanha were both at Yale at the same time. Daniel was my post-doc and Joao was my graduate student. The three of us were interested in switched systems and a lot of work was done with both of them, especially in adaptive control and multiple model adaptive control. Of course some of that work was with you. That was also a very nice and exciting experience. Even more recently I have had a bunch of students from China, who have done excellent work. Ming Cao, who is now in the Netherlands, did outstanding work in consensus and very recently I’ve graduated two more students from China who have worked on formation control. You may remember that you and I got involved in formation control during one of your visits. We were sitting in front of the computer and not quite sure where we were going, which is the usual case with research, and we ran into this thing called graph rigidity. I didn’t know what that was and neither did you. And then it became clear that hey, we ought to be able to use that to set up a structure for doing formation control and that led to an enormous amount of research. And by the way, that continues today.

Anderson:

And that’s one of the oldest topics in terms of being a scientific topic that’s had research one on it in that Euler

Morse:

Yes.

Anderson:

And it’s still got a collection of significant unsolved problems, nothing to do necessarily with formations, just the theory of rigid structure.

Morse:

That’s correct.

Career Challenges

Anderson:

Well, during your time at Yale, you’ve had some great interactions and you’ve had some great students and you’ve faced and dealt wonderfully with some technical challenges and, you know, won prize papers and various prizes and so on. Were there any very difficult challenges, not necessarily technical ones that you, you had to face along the way?

Morse:

Oh, there were many. One bit of advice where that comes up is that for anybody, any young person who might be watching this, don’t fight, don’t fight. There are famous cases in our field of people who maybe they didn’t get properly referenced or something and they got all upset and they started fighting and it became public. The loser is the person who starts the fight because third parties don’t care. To them it’s just a gossip spectacle. So I always advise my students to avoid fights.

Anderson:

I drew one conclusion from your description of optimization. It was as if you were trained to have a huge hammer and to go round and treat everything as a nail. And you realized that wasn’t quite the right way to do things.

Morse:

That’s right.

Anderson:

Yes.

Morse:

And even today, one can be in conversation sometimes with people and the first thing that comes out is ... Oh, I can set that up as an optimization problem... or... Can I set that up as a optimization problem? and you know, that it’s fine if you can do it, but maybe that’s not always the smartest way to do things.

Anderson:

Yes. It is important to allow different views of the same problem, isn’t it?

Morse:

Sure.

Anderson:

Much like an architect’s drawings maybe show a house and elevation and plan view and it’s an ideological era.

Morse:

So let me give a very specific example of this going back to the pole placement theorem. I remember I think this was at Purdue. This was before the pole placement theorem. Some guy came to Purdue who wanted to solve the pole placement problem. So the problem itself, the desire to solve it, predates the contributions of Murray Wonham and Howard Rosenbrock, who also worked on it. This person visiting Purdue set the pole placement problem up as an optimization problem. So you can imagine in your mind’s eye writing down the characteristic polynomial and somehow making a quadratic cost function which of course is very definitely non-convex except in the special case when one has a single input system. For a multi-input system, you have multi-linear form to minimize and this is a hopelessly complicated non-convex problem. The moral of the story is that it is possible to solve the multivariable pole placement problem without attempting to deal with it as an optimization problem at all. Clever people came along like Wonham and Rosenbrock who showed that you could solve this problem easily if you were willing to think out of the box a little bit.

Anderson:

So you were prepared to put the big hammer down and get another tool.

Morse:

That’s correct, exactly. I always give that as an example.

Advice to Young People

Anderson:

So supposing a young person came along to you who wasn’t a prospective student for you and they were finishing their senior year of undergraduate and they said I’m interested in control, what do you think of the idea of going and doing graduate work in control, but various reasons let’s leave you out of the equation. What sort of advice would you give them in terms of where they should try to go, whether that was a good choice of area, how they might pick a problem? What are some of the key things you’d tell them?

Morse:

I think the first thing I’d ask them is do you like to push around knobs and buttons in a lab or do you like to push around a pencil on a piece of paper? I do that quite often, trying to find out if a person has primarily analytical leanings or they’re primarily experimentally inclined. That’s a real distinguishing feature. I think that although experimental work goes on in the control field, it’s a field where at least these days it’s important to be comfortable with the analytical side of things. I would ask if a person is perhaps not so comfortable with the analytical part, I might suggest they move a little bit to the left or the right, maybe do robotics or something like that where you can get more hands on experience. In terms of where, well, the centers of excellence for control in the world continue to move because faculty continue to move and so that’s a function of time. Right now there are a number of good schools. I think Santa Barbara is a place that has a number of strong young faculty, which is where I would be inclined to point people to go. That’s the first place that comes to mind.

Future of Control

Anderson:

Yes, yes, okay. So let’s do a bit of future gazing. Where do you think control will be in 10 or 20 years and, you know, why is it going in that direction or how is going to get there?

Morse:

I’ve thought about this. Let me try to give this answer, which isn’t quite what you’re looking for. So one of the things that came to mind was suppose I had been asked that question 20 years ago. Well, 20 years ago it was 1995. In no way could I have predicted where the control field is now, nor could anybody else, because the control field has undergone a dramatic change. In 1995 we were kind of coming out of the optimal control era, which was a very successful era. The field was looking for something new. In 20 years from now, there will likely be an enormously different than that which exists today. I’m not smart enough to be able to think that far in the future.

Anderson:

And do you think that will arise because of the efforts of researchers or will it arise because companies will be seeking to make money by using certain technologies and to implement those technologies they’re going to need some new control or will it arise because the government is struggling with a number of political questions relating to the economy or health or transportation or something and so it funds special programs and to try to get behind the political problem?

Morse:

It’ll come from all of the above, but especially it’ll come from a few very creative people who just say ”Look, I’m going to ask a different question” and they ask a question, which they can make some progress on and it’s something which has some impact. It’s going to come from very creative people. That’s what drives fields in the end. I think the government plays a role. They certainly played a role in the work, which led to the flocking paper. I suppose the government was a driver behind adaptive control. Industry, too, particularly, maybe some of the commercial products that have come out, particularly now with mobile computing. I don’t know, you know, I can’t say.

Multidisciplinary Control Work

Anderson:

Do you think control’s going to get into bed more in the future than in the past with other disciplines?

Morse:

Yes, absolutely.

Anderson:

And what do you think those might be?

Morse:

Well, computer, computer, computer science, there’s absolutely no question about it. And some might say the social sciences as well. There’s lots of activity involving the social sciences and control. The most recent Control System Magazine has a nice article by a social scientist from Santa Barbara. I’m sure there will be much more of that to come in the future. I’m part of a network science institute here at Yale. Half of the people in the institute are social scientists. The rest are engineers and computer scientists. So we will see more of this, but I always like to put in this caution. The worst thing that one can do is say ”Now we are going to have a multidisciplinary group.” It doesn’t work. So many times I’ve seen administrators make that mistake. What works is people from different disciplines find a problem of common interest and it happens to be multidisciplinary and they go at it. But to try to push people together, even setting up an environment for doing that is a somewhat questionable effort in my mind.

Anderson:

I guess you were an administrator for three or four years at ...

Morse:

I was.

Anderson:

... Yale as a department chair, weren’t you?

Morse:

Yes.

Anderson:

So you’re telling us this is one of the things you learnt then or maybe before?

Morse:

No, I didn’t learn that at Yale. I started observing that as a graduate student.

Anderson:

Yes.

Morse:

Cross-disciplinary research has always been a big buzzword. What’s really needed are ideas and people with a common interest. Well, I’ll give you an example of the kind of thing that didn’t work. There were two fellows at Yale, one a signal processing person who was quite distinguished and knowledgeable about acoustical signal processing. He’s retired now. And there was a guy who did acoustics in mechanical engineering. He unfortunately passed away a few years ago. They both were dealing with acoustical phenomena, but they were on different planets. They asked very different kinds of questions; if there had been a reason for them to come together, fine, but there wasn’t. And they didn’t. I think getting involved in cross-disciplinary work is a little bit like finding a mate. You have to respect the person from the other field, you have to be patient. You have to try understand their language and the way they ask questions, and there has to be motivation.

Anderson:

I guess within companies it’s easier to establish multidisciplinary work because people accept from the start the basic proportion that you do what the company tells you to do. And if you’re asked to join a team to produce some product or create a new service and it’s multidisciplinary, fine.

Morse:

But that’s a little different. Then there’s a goal ....

Anderson:

Yes.

Morse:

... such as to produce a product. People often point to the Manhattan Project as a multidisciplinary effort and it surely was, but there was a clear goal. It was to build an atomic bomb. I think that’s the key ingredient that’s needed that brings people together.

Anderson:

Right. Yes. So you said that major advances will be caused by are likely to be produced by people asking and perhaps solving really new questions or ...

Morse:

Yes.

Anderson:

... getting major new insights, which is to our first approximation I think a solitary sort of an exercise....

Morse:

Yes.

Anderson:

... for a person maybe fed by discussions, but it’s more a solitary sort of exercise, and multidisciplinary teams that maybe have a goal in mind as somewhat orthogonal to that aren’t they? Is there an inconsistency there?

Morse:

That’s a tough question because it’s field-related. I am aware that there are fields where it becomes very difficult to do things as a single individual, but I expect even in these fields, where there is a successful multidisciplinary team there is typically only one or two main innovators. Creativity is the name of the game and that tends to come from individuals.

Career Reflections

Anderson:

I sense that if you look back on your career, there’s very few things you wish you had done differently.

Morse:

Oh, there are so many. I don’t want to talk about them.

Anderson:

Oh, well, that’s ...

Morse:

My mistakes.

Anderson:

Fine, of course.

Morse:

Oh, yes.

Anderson:

Was there any other piece of advice you’d like to leave us with?

Morse:

Yes, I think so. I think I’d like to start with an experience that I’ve had just this past Thursday when I flew down to Washington to attend an Air Force/DARPA meeting. It turned out on the plane siting next to me was a nice lady named Flora Wilson. She was a grandmother. And we just got to talking. She told me about her grandchildren and went on and on. At some point, she said to me ”and what is it that you do?” I said I do automatic control. I’m an engineer that does automatic control. And she said ”Oh, what is automatic control?” I thought about it a little bit. At first I was going to say use cruise control as an example, but maybe she didn’t drive a car. I was in a airplane, so I said okay, the autopilot, there’s something called an autopilot on this plane. It’s flying the plane right now. The pilot is just watching it. And she said ”Well, that’s nice. What’s an autopilot?” I said it’s a computer that makes the ailerons go up and down, it makes the rudder move. It keeps the plane on course, it keeps it from rolling and so forth. And she said ”Oh, you make computers?” I said no, no, I don’t make computers. People in my field specify the rules the program which runs the computer are supposed to carry out. She said ”Oh, you’re a programmer.” At this point I sat back in my chair and I started to think. I was reminded of Donald Knuth’s famous line, which was something like, ”An algorithm has to be seen to be believed.” And I also thought of Paul Hamlos’s famous line and I decided to use it so I simply turned to the lady and said “In truth, I’m in roofing and siding.” So there is a bit of a lesson that goes with this. And that is that the control field is one of the more difficult fields to explain to people who are not in the control field, even technical people. Most of us have tried to explain this to our parents, to our wives, and it’s a difficult thing to do. One way in which people have tried to do this is in terms of applications. There’s a problem with this approach. Of course applications are very important and they are needed to justify a technological field, but most applications of control involve technology which is not control. Even a cruise control system involves all kinds of components and devices, which you can touch, which are not themselves really “controls”. We are a field which deals with things you can’t touch. We act in the ether. And this is part of the reason why we have difficulty explaining ourselves to people out of the field. I think this is extremely important. It affects the perception of the control field by people in industry, the government and academia in terms of funding, in terms of tenuring, everything. And it’s something that I think the control community really needs to think about hard, about how the field might project itself to those outside of the field in ways that go beyond applications that the control community has been involved with. You can say going to the Moon couldn’t have happened without control. And that’s certainly true. But the lunar excursion vehicle certainly was largely not a control system. So let me close with a parallel example. A company, which we all know very well, IBM, divested itself of laptop computers, sold that business off to Lenovo some years ago. I expect that even now if I were to ask many people ”What exactly does IBM do? most would scratch their heads. There is this huge company still making a lot of money. They do make chips and they do make, I guess, some mainframe computers, but they’re not largely in the computer business, at least the hardware business. They’re a service company. I think if you were to look online at descriptions of what they do in the financial world, that’s the way IBM would be described. It’s a service company. Well, we’re a service field. We provide a service. We don’t provide programs. We don’t provide hardware. We provide a service which other people use to make their systems work, and I would say that is an idea which needs to emphasized perhaps a lot more than it’s been emphasized so that people can begin to understand, exactly what it is that we do and what would happen if control weren’t around .

Anderson:

Well, that’s a very good, point. Like you, I guess I’m glad in this sense that we’re not philosophers because I think we’d have a much harder problem.

Morse:

Oh.

Anderson:

Again, it is correct that, one has to be able to sell one’s discipline not just to family and friends, but to politicians and deans or university presidents and senior people in companies. And you’ve made the point that using specific applications is a vehicle for doing that is kind of a aid, but it mightn’t get you there fully in the end. Steve, you’re known, of course, as an eminent researcher, but I think I’m very grateful and on behalf of all of the people who will be seeing this interview at your willingness to have shared so much about your insights about the context in which you’re working, in which you have been working. And I would think many people will get a lot of inspiration from today’s interview and I would like to thank you very much for your frankness.

Morse:

Oh, it was my pleasure, Brian. Thank you for, for going to the trouble of listening to me.