Interview with Minoru Asada

Q1: Could you tell us your name?
Minoru Asada: My name is Minoru Asada. My first name is Minoru, and my last name is Asada.
Q2: We’d like to ask you a question about your humanoid robots research. When did you first begin working on humanoid robots?
Minoru Asada: Conceptually, the starting point of my research from the very beginning was always about humans and robots—maybe 45 years ago or something like that. But as for when I actually started working with real humanoid robotics hardware, it was about 20 to 30 years ago, in the late 1990s, when we started to design humanoid robots and bring platforms into the lab for experimental research.
Q3: What problems were you trying to solve at the time?
Minoru Asada: My line of thinking is that the humanoid is a wonderful research platform to understand what humans are. Researchers across many disciplines—such as psychology, neuroscience, and cognitive science—are constantly thinking about human nature. Designing a human-like artifact deepens and broadens our understanding. In humanoid robotics, I am always asking what humans are, and humanoids serve as a great platform for that pursuit.
Q4: Could you point us to some of your key publications from back then that we should check out?
Minoru Asada: Because I focus on cognitive developmental robotics, I look into the developmental process of how a baby grows and becomes smarter. In English, around 1999 or 2000, I published a paper in IEEE Transactions titled Autonomous Mental Development. That paper was a starting point, moving beyond purely mechanical design toward research on humans and humanoids together. It served as a debut paper for cognitive developmental robotics, though I had published several papers on these issues in Japanese prior to that.
Q5: Was there anything that was difficult in research back in those days?
Minoru Asada: Everything was difficult! When designing a humanoid, biological systems rely on mechanisms like muscles. However, standard humanoid robots rely on electric motors and completely different mechanical setups. We were always trying to imitate biological systems, but due to technology limits, we could only imitate specific parts. There is an endless gap between artifacts and biological systems; once you close one gap, another appears.
Q6: How would you say your research evolved over time?
Minoru Asada: Development itself is very difficult as a research target because researchers usually tackle a fixed point rather than an ongoing process. We attacked individual research topics step-by-step—which is closer to learning than full development.
Over time, I received significant funding supported by JST and collaborated with colleagues like Hiroshi Ishiguro and neuroscientist Toshio Inui. Everyone had their own individual research protocol—for instance, Yasuo Kuniyoshi worked on fetus simulation, Hiroshi worked on social robotics, and Toshio focused on neuroscience. Connecting all these distinct disciplines into a single comprehensive theory of development remains an open challenge.
Q7: What kind of new challenges have you taken on along the way?
Minoru Asada: My primary research question has evolved into asking: What is pain? Pain is a deep topic embedded in biological systems for survival, yet chronic pain and memory-based pain remain mysteries.
Recently, I have been using tactile sensors, artificial neural systems, and brain mapping to explore pain because pain can be the trigger for empathy. First, we look at pure pain inside the robot. Then, through observing pain in others, the robot mirrors that situation internally. That is the beginning of empathy, which leads to mutual understanding, moral behavior, and finally generative ethics.
Models like ChatGPT realize intelligence without a body, which forced us to rethink embodiment. I believe pain is a core component of embodiment; AI language models can express words about pain, but they do not feel it. Embodiment allows an artificial system to feel pain, empathize, and develop intrinsic ethics.
Q8: What technical challenges did you encounter when implementing these concepts?
Minoru Asada: First, soft tactile sensors. A former student of mine designed soft, robust tactile sensors that allow robots to discriminate between soft touch, hard touch, and painful stimuli.
Second, computational models of pain representation. I apply Deep Modality Blending Networks to integrate multimodal inputs (visual, tactile, auditory) into a latent space to minimize prediction error. In experiments, after learning, the robot can predict future pain signals and alter its behavior.
The next stage is mirroring—observing another system, inducing a representation of pain internally, and expressing empathy. Connecting this to reinforcement learning is tricky because physical robots break over thousands of trial runs. To solve this, we use caregiver scaffolding—similar to how a caregiver comforts a baby using language and touch to help the robot conceptualize pain.
Q9: How do you see the relationship between a robot’s body and its controller?
Minoru Asada: I graduated from a Department of Control Engineering, but I don’t view control in the traditional way where a controller is strictly separated from the robot body. Intelligence emerges from embodiment itself; the controller and the body are not separable.
Q10: How does that view change the way you approach robot design?
Minoru Asada: We adapt constantly based on available technology—updating software, feedforward networks, reservoir neural networks, or recurrent neural networks depending on the specific research question. Rather than relying on a fixed control theory, I pick up whatever suitable technology is available at the time.
Q11: Could you give us a brief account of the historical development of humanoid robots and commercialization that you’ve witnessed throughout your career?
Minoru Asada: In the early days, controlling mechanical parts was extremely hard, so the main focus was purely on mechanical design and basic control theory like PID.
I came from a computer vision background. Early computer vision was just simple model-matching, which felt incomplete because it lacked a body. Realizing that physical embodiment was necessary for true recognition led me into robotics.
Over time, materials, electrical components, and operating systems improved. However, controlling humanoids with high degrees of freedom remained difficult. We had to look at neuroscience and nature for the essence of design. Just as airplane developers studied bird flight to capture its essence without copying birds directly, we look at human brains and bodies to extract essential principles for robots.
In terms of key developments, Waseda University’s Professor Ichiro Kato was a major pioneer who thought deeply about human brains and bodies together. Later in 1996, Honda made a massive impact with their humanoid breakthroughs, such as the P2 and P3. RoboCup was also founded with the landmark goal that by 2050, a team of humanoid robots could beat the FIFA World Cup champions. Other pioneers, including Sony with AIBO and QRIO, Washington University, and researchers like Yoshi Nakamura, contributed greatly to humanoid developments.
Twenty years ago, Japan led humanoid robotics. Today, Japanese dominance has declined, while China’s research power and momentum have become extremely strong.
Q12: Why do you think Japan lost its lead?
Minoru Asada: There are two main reasons. First, Japanese government investment is short-term (3 to 5 years), whereas the Chinese government provides massive, long-term funding.
Second, media demonstrations in Japan focused only on successful movements, creating a gap between public expectations and actual success rates. When robots didn’t immediately deliver on exaggerated promises, funding dried up and public disappointment set in. Researchers also shared responsibility for not pushing for sustained, long-term development.
Q13: What do you think about the role of RoboCup in democratizing and developing humanoid technology?
Minoru Asada: RoboCup is a landmark project, much like the Apollo space program. The primary value isn’t just reaching the end goal, but the technological spin-offs developed along the way.
It spans multiple leagues—Humanoid, @Home, @Work, Rescue, and RoboCupJunior. Educating the next generation through RoboCupJunior is crucial because achieving the 2050 goal requires passing the torch. Beyond technology, RoboCup builds an international community involving researchers, local government officials, and industry executives to discuss how robotics will shape future society.
Q14: Do you remember any unexpected failures or successes in your projects or career?
Minoru Asada: In our early reinforcement learning experiments for RoboCup, we ran simulations before applying policies to physical robots. We defined state spaces (e.g., ball distance, position), but during real-world experiments, we realized certain states rarely occurred while others were heavily biased.
Pure theory or simulation is never enough—real-world physical experiments reveal unexpected truths. Research involves endless failed trials before a single success, but failure is simply the process of learning how to improve.
Q15: Is there anything else you would like to add about your current work?
Minoru Asada: My current focus has shifted toward ethics. Embodiment combined with language interfaces makes it much easier to deploy robots in society, but we must consider both positive and negative societal impacts.
For example, anthropologist James Wright interviewed care home residents in Japan about care robots. He observed that government and industry initiatives often focus strictly on utility (e.g., lifting senior citizens for baths). However, residents expressed feeling treated like objects rather than humans when interactions were purely functional.
We cannot design for raw functionality alone. We must adopt agile governance—bringing together designers, users, caregivers, and local governments from the very beginning. I now teach ethics, virtue ethics, and intentionality to undergraduate students so they design with societal impact in mind.
Q16: Do you have any advice for students wanting to get into the field today?
Minoru Asada: Please come to robotics and design robots as you like! But always remember: your robot will be used by diverse people across different countries, cultures, and modes of interaction.
Think about what makes us different and how we interact. By building empathic capabilities into robots, they become trustworthy companions. Focus on how humans and robots can communicate and coexist harmoniously.