Interview with Bernie Roth, by Micah Lande, November 2006

Bernie Roth first became interested in kinematics during his graduate work at Columbia University. He entered robotics – nearly a nonexistent field in 1964 – when a computer science professor heard about Roth's mechanical design knowledge and asked him to help. As Roth explains, that fateful moment of somehing thrust into his lap changed forever.

LANDE: So your primary field of academics is robotics and kinematics. Tell me how you got involved in that or how you came across it.

ROTH: Kinematics was the subject I originally taught at the City College of New York while I was doing my graduate work at Columbia University and I became interested in that. There was someone at Columbia who had made a big splash in the field as a young man, and I sort of naturally gravitated towards him and became his second Ph.D. student. So that's how I go involved in kinematics. And when I came to Stanford I was working in kinematics basically, and then I got a phone call from, after I was here for a year or so or two, I got a phone call from Frederick Terman, who this building [Terman Engineering Center] is named after, who was the provost at the time and he said someone had just talked to him and this Professor McCarthy in Computer Science had gotten this large grant, I think it was 1 million dollars, which wouldn't be considered large at the moment, to work in robotics, and they were concerned that he didn't have any mechanical people around him who knew about designing things and they felt I should talk to him. And shortly thereafter I got a call from McCarthy and it got started off. So it wasn't something that I initiated, it was something thrust upon my lap and I was responsive to it and it turned out to be a big part of my work.

LANDE: About what time is this?

ROTH: Oh, I would say 1964.

LANDE: I would imagine robotics and kinematics were relatively new fields then?

ROTH: Well, kinematics is a very old field. Robotics was a non-existent field at the time. There was basically no work when I started to look into it. There was one Master's thesis at MIT where someone had picked up a block using some sort of a mechanical arm thing and that was all. So we were relatively early in the field.

LANDE: Could you define kinematics for me then–is that the study of movement?

ROTH: Kinematics is the study of motion, usually without regard to the causality, which is part of the game. So kinematics is basically the geometry of motion and it could be applied to various things.

LANDE: The name Edwuard Muybridge comes to mind. His early study of motion supported by Leland Stanford?

ROTH: Muybridge–you're talking about the photography stuff? That's at Stanford actually, as an aside, when I came here I tried to look at some of that stuff and they had the nude men swinging baseball bats hidden away as though it was pornography, that's changed. [Muybridge] is regarded as the history of photography really, and there is some interesting kinematics, but none of that really related to kinematics directly. Kinematics is more the study of the geometry of motion. Muybridge's work with a horse galloping and all that has to do with dynamics because a horse can certainly pick up all four legs so they could or could not be off the ground. Kinematically there is no constraint that way–the constraint is more in the forces and things like that. That wouldn't really be central in kinematics in those days.

LANDE: Working with the computer science professor, and the other projects that you were working on at the time, what were some of your favorite projects or some of the interesting problems that you faced or tried to address?

ROTH: My interest right off the bat was to make a science out it and to not just do ad-hoc building of stuff. So there were two demands that people running the project from computer science hadtheir interest was to just get stuff, demos and things like that. So to satisfy that we got a hold of a prosthetic arm from Rancho Los Amigos hospital and instrumented that and put it under computer control and then we designed a hydraulic robotic arm that was very fast. The belief in those days was that the computers could do everything and you didn't have to worry about the mechanical design. It was easily fast enough, so you just give them anything and the computers will take care of it and everything will perform well. That wasn't true and it still isn't true and the design becomes very important. That was the belief to begin with. So we tried to build a very fast hydraulic arm, which we did. Eventually the reality was that you didn't need arms that fast, the computing was that much more complex, and so we ended up building an electric arm, which was called the Scheinman Arm, which was a thesis that Vic Scheinman who's still around here did as an engineer's thesis basically.

LANDE: What was your involvement with Vic?

ROTH: He was officially my advisee. Then I took sabbatical so actually Ken Waldron, who's back here now who was my second graduate student, took over, so essentially Vic has been my life-long advisee.

LANDE: I've heard a bit about the VicArm from Vic himself.

ROTH: We did all that and that was important to get things going and made a lot of movies and demos, but at the same time it was a doctoral thesis that set the basic science for the field in terms of how you categorize these things, how you derive equations to design them and control them, and things like that, how you classify them. Originally people just had classifications based on brand names, like saying a Mercury car and a Ford car and this car, without looking at the commonality between those. AMF (American Machine and Foundry) had the first one and then Univation had one and then eventually a bunch of Japanese came into the field and a few Europeans came in. But we tried to establish a commonality without worrying about the brand name but instead more about structure, to talk about what parameters are important. So those early theses kind of set the tone for that. A bunch of books came out of that and defined the way the field is kind of done now. If you look at the manipulator part–there are other aspects we didn't work in that have to do with vision, which other people are doing in the laboratory, planning and things of that nature.

LANDE: You mentioned an assumption that it seems people made early on, that the robotic arm could do things better than a human could do them. Were there initial assumptions that with an optimistic view of technology provided a resolution of that?

ROTH: The first arms were so bad that the amazing thing is that they did anything, really, rather than better. More than anything, it was the people who had a commercial interest in this stuff that were selling things like safety, so that was how it was sold originally. And they would use the original arms that went into companies when it was an automotive company and they were used for things like big punch presses and things like that. So they were in factories, so it was sold that way. Amusingly enough there were provisional robots, hands that were totally crushed in the presses the same way human hands would. Of course, it was better that it wasn't human arms. They were a number of early industrial accidents to these robots. But the basic idea was you get people out of dirty dingy places and things of that nature. People didn't talk about the economic aspects as much as we talk about it.

LANDE: With a lot of the actual widespread use of these arms, I'm thinking mainly of the automotive industry in the 80s, at least in the United States, there was a backlash towards machines replacing people in these jobs that require routine. Was that a consideration in robotic science? Or was there a catalyst for that type of conversation to happen in academia?

ROTH: The conversation never happened. There is always the counter argument that these things create more jobs that they take away. Whether that's true or not depends on how you do the arithmetic. The way that's handled in general is, in industry, when they want to bring in these things, even numerically controlled things, they use it as "there's always progress" and "if you don't do it, we'll lose the company" and that kind of thing. And if there isn't any resistance in general, they buy off the current workers and send them away and do not worry about future generations. It is a conversation that took place. It was never a very strong conversation in America I think because basically the union movement isn't as strong as it is elsewhere. I think it was a stronger conversation elsewhere in the world. It didn't exist in Asia because there is no movement to speak of company/unions in Japan basically. In Europe I think there was more of a conversation about that. I don't think it was ever a strong thing.

LANDE: The cultural aspect of it is interesting. What's the difference between Japan/Europe/here with regard to robots?

ROTH: In Japan people have written books as Japan �the robot kingdom,� and it is true that there is more of a fascination for robots in Japan than anywhere else in the world. And that's true with culture in terms of selling toys and the [Sony] Aibo-s and those things like that. But in terms of factories, I myself, in going to visit automobile factories around the world, haven't seen much of a difference between Japan or America or Europe, or at the least it's the same devices and controls so far as I can tell. And even assembly lines and things like that. There are a lot of robots in Silicon Valley doing biology and the controls so far as I can tell essentially are done the same way as they are in Europe.

LANDE: When I think of robots in industry I think of an assembly line and cars. You mentioned biology, which to me that seems like a new idea, maybe it's been around for a while. What are the normal/abnormal applications?

ROTH: Well, you go to a place like the Stanford Linear Accelerator and they have robot arms. Any place where there is a repeated scientific experiment where you have to move stuff from one place to another or mix something and it has to be done over and over and over again you can replace it with a machine. Now if the machine looks like a mechanical arm, you call it a robot. The truth is that these are a lot of machines that are called robots that probably shouldn't be called robots in that sense. They are just fixed machines that do the same thing over and over again but don't use any intelligence or anything like that. And most machines in industry are that way. But in fact, there is a use for things like pipetting and biological testing and manufacturing that involves repeated motions, and that's basically where you could use mechanical arms for these machines that do these repeated motions. But it's true that the big thing that we know historically as the industry that was first user and then got into assembly and being in packaging, but many places they use it, UPS uses it. Anything that is a repeated start/stop or from here to there you can do with a mechanical arm.

LANDE: The term arm for me is rather evocative of a joint with 3 degrees of movement. Is there a standard definition that you have? Of the mechanics of how it works?

ROTH: It's kind of funny because for a long time I didn't use the human-like words like "arm" and "hand," I used words like "end-effect" and "manipulator." I think those words are better words to use, but it's a losing battle. It's true that many things which are mechanical manipulators don't look like human arms and they have different geometry and kinematic structures. Some do. We tend to generally speak loosely and use the world "arm" and the word "hand" even though they don't resemble the human arm or hand in any shape or form.

LANDE: The basic function is not based on any observations of biological?

ROTH: Not necessarily though. There's no, what they call, biomimetics. There is very minimally in this manipulator stuff and end effector stuff. There have been some people who have tried to replicate the human hand and the arm and especially people who do prosthetics do that. And that's a field very closely related to robotics– in some cases it is the very same people involved in both. There are some examples of all that. In general, if someone just designs a mechanical manipulator or arm, it has nothing to do with the human arm. Putting something at the end of the manipulator to grab something has nothing to do with the human hand.

Copyright 2006 Ambidextrous Magazine, Inc.