Interview with Masato Hirose -- "Falling down, getting up, and walking on"
Interview with Masato Hirose
Manager and Chief Engineer of ASIMO at Honda R&D Co., Ltd.'s Wako Research Center
Hirose is the inventor of ASIMO, a humanoid robot launched on November 20, 2000, by Honda Motor Co., Ltd. He was engaged in the design of automated drafting instruments and other items at a machine tool manufacturer, but left the company to join Honda Motor Co., Ltd. in July 1986. Shortly thereafter, he became engaged in robot development, a field in which Honda had only just started research at that time. From around 1990, he played the leading role as the group leader in the robot development. Born on February 7, 1956, Hirose is 46 years old. His home city is in Tochigi Prefecture ASIMO is 120cm in height and weighs 43kg, the size of an elementary school first-grader, small enough to operate in the living space of an average household and to encourage a sense of affection towards the robot.
--In autumn 2000, two humanoid robots - Honda's ASIMO and Sony's SDR-3X - were released consecutively, giving one the feeling that a new age was dawning. Were there any exchange of information and personnel between Honda and Sony on the robot development?
Hirose: Yes, there were some interchanges. In fact, I even had a chance to visit the research office of AIBO. I had seen sketches of humanoid robots in some of the materials Sony showed me earlier, and I had thought that it wouldn't be long before they would come up with a humanoid robot. My impression on seeing Sony's humanoid robot was that they had finally done it.The movement of Sony's robot didn't particularly astound me, but I must admit that I was quite surprised when I saw it break into apara-paradance. I was amazed to see the robot move so quickly and smoothly. Anyway, I certainly do welcome robots debuting from other companies.
--What brought you to begin the development of a humanoid robot?
Hirose: I applied for Honda's offer of employment I saw on a newspaper and joined Honda R&D Co., Ltd., in July 1986. On my second day at Honda, the managing director came up to me and told me that I was to work on robot development. Honda is a company of mobility, as in motorcycles and automobiles. Knowing this, I had naturally assumed that I would be in charge of developing these kinds of products, and so I was quite surprised to hear that, to be honest. On top of that, my mission was not to build a simple robot, but what they had called an "Astro Boy" - looking back, I think they gave the name probably to set distinctions to what they wanted me to develop was a humanoid robot and not a robot for industrial use. In this sense, I do not think they wanted me to develop a real "Astro Boy" word for word.
--Were you any way perplexed when you were told to develop humanoid robots?
Hirose: No, not especially. Before joining Honda, I was engaged in designing automated drafting instruments at a machine tool manufacturing company. Thus when I was told that I would be working on robot development, I thought it would be something similar to what I had been doing, because be it an instrument or a robot, it is something that moves. Honda had only just recently started research and development of robots at that time, and the only thing in our office were some desks, chairs and several books. We started off with an entirely clean slate.
We must take the lead
--What was Honda's objective in developing a humanoid robot?
Hirose: The underlying hope was that conducting robot R&D would enable us to accumulate such kinds of technology as control technology, environmental sensing technology and material technology. No one would talk about selling robots commercially or set this field as one of the company's profitable businesses. Our team's goal was to conduct basic research, and the top management attentively made arrangements so that we would not feel constrained by money worries. Mr. Kawamoto [Nobuhiko Kawamoto, Honda R&D Co., Ltd. President and CEO from 1990 through to 1998] once told me, "Researchers shouldn't have to think about money. Instead, they should devote themselves in research and development." To this end, the company would only give us the proposition to develop "Astro Boy" and freely let us work out all other tasks by our own will.
--Which do you think is more of a blessing for engineers: Being involved in a basic research project for the future or working on a project that would directly link to business?
Hirose: Both are rewarding for engineers. At the former company I worked for, I experienced to sell the product that myself had manufactured. Of course, it is quite thrilling to be able to introduce to the world something that myself had devoted so much time and energy in the development.On the other hand, it is also a great joy to assume responsibilities from my own company for working on something like robots that few other companies have embarked on. Such opportunities do not come along these days. Above all, I adhered to the idea of making a two-footed robot walk. What kind of research other companies and universities were conducting was almost not a matter to me. By all means, I did read and refer to books from time to time, but doing exactly what someone else has done before is like retaking an exam. You cannot move ahead of someone by just following his footsteps, because you would not be able to see the road ahead. "Hold the torch yourself" is a saying that my seniors at Honda taught me -- you cannot create new things unless you take the lead
Falling means failure
--Your work may be fun, but all the more, you have to produce results too.
Hirose: Exactly. For instance, automobile races are often referred to as "running laboratories." It is okay when you win and attain the first seat in the race, but slipping into the second place means being the second best. In other words, you did not win. Results are everything. There are times when the process of the race is reviewed, but that is not evaluated. It is the final result that counts. Same thing can be said for robot development. To date, we have developed various element technology and mechanisms. Each time we develop something anew, we show the robot to the company's top management. They tell us to make the robot walk and we make it walk. But the robot falls down. We hear from the executives, "It is no use if it cannot walk and falls over." That is it. It does not matter how well it walked the day before if the robot falls down the day we show it to the audience. And much to my consternation, it seems that the only time the robot falls down is the day we show it to the company's top management. We feel utterly dejected. Obviously the robot falls down because the technology is not sufficiently mature. The only concern of the company, though, is whether it walks or not. Over the course of robot development, we do end up in devising a control technology and obtain a patent, but such was not what the company wanted from us. What they wanted was a humanoid robot. It was as simple as that. Given that only results count, the process itself did not matter at all. It never really occurred to us to hand in technical papers; only once in the 14 years of ASIMO development did we present any papers. Maybe obtaining patents was the only thing that was necessary other than developing robots. Registering patents and creating a humanoid robot were the two tasks for us, and we devised every means to accomplish them.
All were fervent
--Weren't you under any pressure with the development taking as long as 14 years?
Hirose: Well, of course, we were not given 10 or 20 years for R&D from the beginning. R&D period for automobiles is four to five years. Initially we were told to develop a humanoid robot within three to four years, but we ended up taking 14 years, so we were under some pressure, to be sure. Still, we never thought of giving up. It was because all of the members of the R&D team were diligent and passionate about their work. We would ambitiously set high goals and because of this, we would always look for ways to attain that goal. The team consisted of a string of these kind of eager researchers. Thus we never had a happy-go-lucky atmosphere in our office. Nor did anybody saunter the room wandering what to do each day.
--So the team was functioning well.
Hirose: During the first few years of development, we had a team of a dozen or so people who were all mechanical engineers, and no one ever had to face a difficult technical problem alone. I think we had a culture where we could speak out what were in our minds. We would get together everyday, anywhere, for "bull sessions." We would discuss in a raucous and an up-front manner, and not necessarily were we forced to reach a conclusion; we usually bounced ideas off each other, sharing our thoughts on doing this or trying that. After a day or two of holding talks, surprisingly though, we would always come up with ideas on what to do next.By all means, we would discuss matters thoroughly whenever we had to make decisions for setting technical directions. A couple of hours would soon fly even when our topics were on trivial issues. Let me give you one example. It was when we were trying to move on to the next stage of development from "E6" -- a torso-less prototype -- to "P1" -- a test machine equipped with an upper body. I suggested to my members that if we were going to attach an upper body, we should ensure that the robot could freely move in the waist area. However, everyone else was against my idea. Although I was the development team leader, I was the only person to insist. As I could not concede to their idea, I mused upon the matter all the way home from work. I got off at the station and walked around for a while until I happened to look around. I could not recognize where I was. I said to myself, "Huh? Where am I?" and then I realized that I had got off one station too soon. As such, developing humanoid robots totally engrossed my mind. I suppose everyone else was pretty much feeling the same passion for robots. In the end, I gave up the idea of giving the robot a freedom in moving its waste area for that issue.
A brisk walk to the finish line
--Around which period did you start finding difficulties in the robot development?
Hirose: We had a hard time getting the robot to walk at the same speed as a human being. In just the first two to three months of development, we managed to get the robot walk as what we dubbed "static walking." But the walking speed of the robot was extremely slow, not even reaching the 1 km/hour-level. Enabling the robot to walk in a "humanoid" manner meant it had to walk briskly at a speed of about 5 km/hour, "dynamic walking" is the word. We finally attained this goal in around 1988-1990; the walking speed of the experimental "E3" model went up to 3 km/hour and the "E4" at 4.7 km/hour. Thus the tough time for us was around period 1986 -1990. During these five or so years, we devoted ourselves in observing how people walked. From time to time, we videotaped and examined people walking and even attached lights on various parts of people's legs to study how their legs were moving when they walked. At times, we visited and questioned orthopedic surgeons on how the human shoulder and pelvis move and how to take rhythms when walking. We even studied people in the context of day-to-day living, watching people walking up and down the stairs of train stations.
--During such period of trial, when did you feel certain that you could make the robot walk?
Hirose: I cannot indicate the exact point in time I felt that kind of certainty. Our R&D approach was to imitate the walking style of human beings. We wrote a computer program on walking style, transplanted the program into a mechanism, and then tried to get the machine to walk. When it fell over, we carefully reviewed what happened and re-programmed the mechanism and it was a repetition of trial and error. And we reached our goal of speeding up the walking speed of the robot to 4.7 km/hour, as a result of these trials. It was not something that we tried and accomplished in a flash.Having said this, I suppose we inked important milestones when we completed the walking mechanism with the "E3" in 1991 and when we developed the technology that prevented the robot from falling over with the "E6." The body was a bit large, but we were proud of the fact that the E6 looked like a robot.In that sense, I felt that we had crossed the finish line when we wrapped up our work on the "E6". Our company is quite extraordinary, though. As soon as we saw our completion of the E6, the top management said, "We told you to make a 'humanoid' robot." Since we had gone to all the troubles of getting the machine to walk on two legs, we were told to give the robot hands and to make these hands move. I was thrilled to hear this. The thought of being able to continue developing the robot made me feel excited. After all, it is great fun. I enjoy working on the robot.
Bring me a can of beer
--What kind of things would you like the robot to be capable of?
Hirose: When people ask me what kind of robot I was aiming for, I would always reply that I want a robot that can go get me a can of beer from the refrigerator. This may sound simple enough, but in reality, it is quite difficult. There is quite a number of hurdles to overcome: Enabling the robot to figure out to which side the door of the refrigerator should be opened and to determine what item to select from the vast arrays of items inside the fridge.I did try once in the past to create a hand of the robot. I observed how people use their hands, and after all that study, I gave up. Human beings use every single finger of their hands. I learned that people use everything they have. That is the reason why I gave up. The task of creating a human-like hand for robot was just too daunting.It is extremely difficult trying to get a machine do all the things that the human hand can do. So currently, my focus is on developing more basic operations for the hands of the robot. Surprisingly enough, people do not usually carry hefty things around in their day-to-day living.
--Once you developed a robot that can bring you a can of beer, what is your next objective?
Hirose: I think that a robot is simply an extension of a machine, so I intend to make it a tool that is convenient. That being the aim, I do not see any necessity in adding too much intelligence capability into the robot. I am not thinking of trying to give the robot the ability to make its own judgments. What I have in mind is something akin to a housemaid -- A robot that will bring something over to you, do the shopping for you, and so on.
--So it seems your concept of a robot is quite different from that of Sony's biped humanoid robot. Sony has clearly stated that there was no need for a robot to be useful in terms of serving humans.
Hirose: Sony's robot is serving the humans by soothing people's heart. The difference between our robot and Sony's is that ours is meant to be useful for humans as a machine while Sony's robot appeals to people's hearts. However, what we must bear in our minds is that even when we are developing robots as useful machines, you cannot disregard their interaction with human beings. People often become attached to their cars. Certainly the car is a very convenient machine, but the driver is the one who sits and conveys his/her wishes to go left or right to the machine by taking hold of the steering wheel and drive. The same thing can be said of a cook looking after his own knives carefully. Making a robot move in accordance with your wishes might also give you a sense of attachment to the robot or the robot itself might feel some kind of affection to you. My idea is that the more often you use an item conveniently, the more you would be feeling a certain attachment to it -and I intend to develop a robot based on the equation.