Interview Transcript
Q1: Professor, thank you so much for your time today. To begin this interview, could you tell us your name and a bit about your upbringing?
Eiichi Yoshida: So my name is Eiichi Yoshida. I was born in Tokyo and I grew up around the Tokyo area. My father is also a researcher engineer in New York City. Then I entered the University of Tokyo and also studied engineering.Then I had a chance to go abroad, to Switzerland, EPFL, as an exchange student. That gave me some opportunity to learn French. That may later play a big role, but I will tell you later.
Q2: So tell me a bit about your graduate school experience.
Eiichi Yoshida: I started working on robotics in graduate school, during my master’s. It was the control of robot systems, including flexible arms. Then I completely changed my research after returning from Switzerland to multi-agent systems. A multi-agent system is where you have autonomous agents that collaborate or cooperate. I worked on the local communication of those agents, including how to establish communication and how to make cooperation between them. I did simulations, and this distributed system became the topic of both my graduate thesis and my doctoral thesis.
Q3: Why did you switch topics after your experience in Switzerland?
Eiichi Yoshida: Because the student after me completed the work I was doing, and I was also deciding whether to continue to a doctoral course or start working. My professor, who was my supervisor, recommended that I continue. In that case, I needed to explore something new. Also, in our laboratory they had started working on multi-agent systems, so I decided to work on this kind of distributed multi-agent system. This was in the 1990s.
Q4: Could you tell me how you got interested in robotics in the first place?
Eiichi Yoshida: When I was a kid, there was a famous cartoon called Doraemon. I was very fond of this cartoon. In this cartoon, Doraemon is a kind of character between human and cat, but he has very human-like characteristics. He came from the future and helped a child. He takes out many tools from his pocket, called the fourth-dimensional pocket. The tools he takes out are very interesting, coming from the future, like understanding people, complete translators, teleportation, and so on. I really liked these kinds of stories, especially the technology.
Q5: Would you say you are still influenced by the idea of Doraemon to this day?
Eiichi Yoshida: Yes. These future technologies have actually become reality, such as translation technology, and I think this is very interesting. I generally became interested in technology through this cartoon. Also, there were many robot animations in Japan at that time, like Getter Robo and similar shows. These robot animations were also a good source of my interest.
Q6: Do you remember the very first robot that you saw or heard of?
Eiichi Yoshida: Since robotics research started in Japan when I was in the fourth grade, I was assigned to this process. There was a big industrial arm, so this was the first robot I became familiar with. My first robot was a very primitive one. It had two flexible arms driven by motors. This was the first robot I really controlled.
Q7: Since the Doraemon days, you mentioned that you have been interested in technology, but what have been the main research questions you have tried to answer in your research?
Eiichi Yoshida: Control has always been something in my mind. But after working on distributed systems and multi-agent systems, I became interested in how many robots can collaborate and work together to be more efficient than one robot or a centralized system. This was one important question. I continued similar research after joining the Mechanical Engineering Laboratory, which was a public institution in Japan. There I started working on self-reconfigurable distributed modular robots. This was also a continuation of the research question about how autonomous or semi-autonomous agents can collaborate to perform tasks together, especially how a self-reconfigurable robot can change its shape according to the environment.
Q8: You have had a very successful academic career, but have you faced any challenges?
Eiichi Yoshida: I don’t know if it has been successful, but I have experienced encounters with new people who influenced my decisions. Almost every ten years, something happened. The first one was when I had been working on self-reconfigurable modular robots. The director of the institute, the Intelligent Systems Research Institute of AIST, which was reorganized after combining several public laboratories under the Ministry of Economy, Trade and Industry, Professor Tani, suddenly asked me: “Would you like to go to France? We will start a collaboration, but the topic may be humanoids.” This was completely different from what I had been doing at the time.
Q9: Could you explain that collaboration and why it was such a challenge?
Eiichi Yoshida: They were going to start a joint robotics laboratory between our institute and the French National Centre for Scientific Research, CNRS. The unique feature was that there were two sites, one in Japan and one in France. There were two co-directors, one French and one Japanese. In Japan, Professor Tani and Dr. Yokoi were the co-directors. I went to France as the Japanese co-director, and my counterpart was Dr. Jean-Paul Laumond. Unfortunately, he passed away several years ago. This was a big challenge because my wife was working at a company, and my children were around one year old and six years old. It took me a long time to decide. But I thought, why not take on a new challenge? So I left Japan to join this joint laboratory in France. It was challenging because I had almost never worked on humanoids at that time. There were also cultural challenges and challenges working with French institutions.
Q10: What were some of the cultural and technical challenges of introducing humanoid robots to France?
Eiichi Yoshida: At that time, humanoid robotics in Japan was leading. Twenty years ago, Japan was very advanced in this area. The topic of this joint laboratory was to increase robot autonomy using a humanoid platform. So I needed to introduce humanoid robots to France. I was afraid that, because of cultural differences, humanoid robots might not be accepted. Reproducing humans with robots is a very delicate issue. There was also a funding problem. I needed to work together with Jean-Paul to find funding to introduce the robot. At that time, it cost around €400,000. People told me, “You gave this robot to France.” No, no, no. The French institution bought the robot. Fortunately, Jean-Paul found the funding, and we introduced the robot. Then we had the same platform in Japan and France.
Q11: How did the collaboration between Japan and France work?
Eiichi Yoshida: In France, especially, theoretical robotics was very advanced, particularly motion planning. In Japan, we were very good at control and humanoid robots. So it was a good mixture. They were very happy because they had a very impressive platform to apply their theories. I think we had a very successful collaboration. I worked with them to develop whole-body motion planning for humanoid robots.
Q12: Your French worked well in that context. How did that happen?
Eiichi Yoshida: Why do you ask? Because when I entered the interview, maybe he remembered that when I joined the Mechanical Engineering Laboratory, he was there and maybe I said something about French. He thought, “Oh, I know someone who speaks French, so you can go there.” Not that easy, but in some way it happened like that.
Q13: The humanoid robot back then was €400,000. How did you convince them to purchase it?
Eiichi Yoshida: Fortunately, he had funding that could be used for general purposes at that time. It was probably funding from a company, but he had some freedom in how to spend the money. Also, Japanese robots were not as weak as they are now. That helped. I was also afraid that humanoid research might not be accepted in Europe. But they were familiar with Japanese robot animation. They called it Goldorak, which is the French name for the Japanese robot animation Grendizer. Many people were not against humanoid research. I think this was one of the first humanoid robots introduced to Europe. We even created a new group in the laboratory for humanoid-related research. The group name was JRL, which was Jean-Paul’s idea.
Q14: What makes a humanoid robot such a great platform for research?
Eiichi Yoshida: A humanoid robot has many degrees of freedom. When you control the robot, you need to control these many degrees of freedom, so you need new ideas for controlling redundant systems. Unlike a robot arm, it is not fixed to the ground, so you need to think about dynamic balance. Also, because it has many degrees of freedom and limbs, it can manipulate large objects. You can think about many interesting ideas. Motion planning is also interesting because humanoids have more possibilities to perform different tasks compared with robot arms. That is why people are very interested in humanoids.
Q15: What applications are you thinking about for humanoid robots?
Eiichi Yoshida: Humanoids can adapt to environments designed for humans. When I was working with the joint laboratory, there was a project with Airbus to help humans with airplane assembly. Airplane assembly is very different from car production. In car production, the car moves between robots. But in airplane assembly, humans enter the product and perform many different tasks. Humans are forced into constrained postures and repetitive tasks. Humanoids can have an advantage because they can access these human-designed environments. The floors are not always flat, and there are many obstacles. Also, you have CAD information about the product, so you can perform more precise tasks.
Q16: Why do you think humanoids are not yet common in homes or workplaces?
Eiichi Yoshida: There are several factors. First is environmental recognition. Robots need to understand their environment to perform tasks. Even with advances in artificial intelligence and learning, there are still many challenges. Another factor is cost. It is amazing that now you can buy a humanoid robot for several tens of thousands of dollars, but I think it is still expensive. Maybe if it becomes cheaper than a car, that could become a threshold. Robustness and versatility are also important issues. I think it may take some time before humanoids are introduced into homes, but industrial applications are accelerating.
Q17: What are you most excited about regarding humanoids today?
Eiichi Yoshida: The exhibition at this conference was amazing. I cannot believe how many companies are producing humanoid products. They are becoming very robust and agile, which makes me believe the future is promising. Also, many different learning techniques are starting to be applied to humanoid bodies, including balancing and manipulation. Twenty years ago, Japan was leading humanoid research, but AI and learning were not ready. Now, with recent advances in artificial intelligence, we can expect major progress, especially in industrial scenarios.
Q18: How soon do you think humanoids will appear in industry?
Eiichi Yoshida: I think in five years, we will see some industrial scenarios.
Q19: What would you recommend to students interested in robotics?
Eiichi Yoshida: I think traditional engineering and fundamental robotics technology are still very important. They provide a strong foundation for research. Students should first learn robotics based on models: basic model-based robotics. Then they can move toward learning methods and physical intelligence. Mathematics, physics, control, dynamics, and robot models are important. These fundamentals give you a good basis to grow your research. There are many young people excited about robotics, and I think robotics is a very good foundation for the future.
Interviewer: Thank you so much.
Eiichi Yoshida: Thank you.