Studying humans through robots - a chat with Prof. Asada
Professor Minoru Asada of Osaka University in Japan is a leader in the field of “Cognitive Developmental Robotics” which aims to understand the development of human intelligence through the use of robotics. He is well-known as the director of the JST ERATO Project on “Synergistic Intelligence.” (Think CB2 – which stands for Child robot with Biomimetic Body.) Additionally important facts about him are that he is the vice director of The Japanese Society of Baby Science, an academic group of researchers studying infants, and is one of the founders of RoboCup.
His goal is to understand humans through robotics and then utilize that knowledge to design a robot that can learn by itself. Now that the 5 year ERATO project is coming to a close this spring, GetRobo asked him about what he wants to do next. (The interview took place in San Francisco during Prof. Asada’s visit to the Bay Area in the end of Nov. 2010.)
Q. What are the achievements of the Asada ERATO project?
A. There are many but let me go over some of the highlights of the research done by the 4 groups involved in this project.
First of all, Prof. Koh Hosoda’s group set out to prove that it’s not just the brain that is controlling human movements but that other parts of the body are computing as well. They developed robots that mimic a 7 month old and 13 month old baby using pneumatic muscle actuators and showed how the spine was playing an important role in controlling crawling.
Prof. Yasuo Kuniyoshi’s group was able to show through computer simulation how a 25 week old fetus is able to learn about its body by moving its muscles and touching itself inside the womb. We think that in real life this helps the baby in organizing its movement right after birth.
Plus, we now have constructive proof that a neonatal infant can obtain an image of his own face without any visual data but just through touching. This work was done by Prof. Toshio Inui’s group.
Last but not least, and perhaps the best known outside of academia, is the research done using the CB2 (pronounced CB square) robot, which was developed by Prof. Hiroshi Ishiguro’s group. They taught CB2 how to stand with the help of a human pulling it up. Through this experiment they figured out how to determine whether an attempt to stand was successful or not so that CB2 can improve on its skill.
The common thread in all these research is that we were able to move one step closer in understanding how we can develop a robot as close to a human being as possible.
Q. The CB2 was called “the creepiest robot ever.” What do you think about this?
A. We take it as praise because it probably means that the robot looks real. But it’s actually not as eerie as people think. When you actually see it, it’s rather cute. I know reporters who’ve come to see the robot and commented that it’s cuter than they had thought. The photos and video don’t do justice to it.
Q. What are the future plans for CB2?
A. Initially we had planned to use it long-term and to integrate the functions that were developed with other platforms, but unfortunately due to some hardware problems, we haven’t been able to do that yet.
Q. So what’s next?
A. So far we’ve developed 7 robotic baby platforms. The neonatal M3-Neony, Nana-chan (7 month old infant), Noby (9 month), Hitomi-chan (13 month), CB2 (1.5 years old), M3-Kindy (5 years) and M3-Synchy (5 years). We’ve used each platform to conduct research on each stage of a human’s life.
Next at my lab, we would like to develop a new platform that can be used to do research spanning all these stages. I want to find out about the principle of how humans start to perceive self from others and the mechanism of social development.
The new platform is named Affetto. Affetto can show various emotional expressions and therefore make it possible for human caregivers to interact with it naturally. This is very important to model the early social development of humans.
Q. I read that you are planning to sell the M3-Neony and M3-Synchy platforms. How much are they going to cost and who do you think will be interested in purchasing them?
A. M3-Neony will cost 3 million yen and the M3-Synchy 800,000 yen and they will be sold by Vstone. Vstone is planning to develop the necessary software in about a year. Our intention is that cognitive scientists and psychologists will find them useful.
Currently, we are conducting a joint research where we are having children with Asperger’s disorder play with the M3-Synchy.
Q. When do you think that a self-learning robot will become true?
A. That’s a hard question to answer since it depends on what you want the robot to learn. For example, if we suppose that the learning target is vowel imitation, to a certain extent it’s already possible by preparing the right environment.
But in a real life situation, the robot must deal with various kinds of issues, including the complexity in processing auditory, vision and other sensory information. The many degrees of freedom that the humanoid robots have make the problem even harder. Cognitive functions are inseparable from these sensori-motor issues.
Having said that, my hope is that our quest for the design principle of cognitive development will enable us to realize general self-learning robots in 10 years.
