Robot Mimics Both Humans and Monkeys 2009/10/08 Improving the brain machine interface so that robots understand human intentions is essential in expanding practical applications of robots. The main trend in BMI has been to transmit information from the human brain to the robot’s computer, but the research group of Dr. Mitsuo Kawato, director of the Computational Neuroscience Laboratories at the Advanced Telecommunications Research Institute International (ATR), has developed a robot that can learn by imitating humans.

Flexible joints are another characteristic of the CBi. Gear mechanisms are used in most robot joints, but the CBi’s main joints use hydraulic cylinders. This enables smoother and more human-like movements than in conventional robots. The CBi is also equipped with high-level posture control with the use of a three-axis gyroscope and acceleration sensor, so that it keeps its feet and does not fall over easily even when pushed or shaken by humans.

The CBi is also at the leading edge of BMIs in which the computer understands brain activity. Its abilities were demonstrated in January 2008 when it showed that it could imitate a monkey.

In a joint experiment between ATR and Duke University in the United States, Duke Professor Miguel Nicolelis and colleagues had a monkey walk on a treadmill, and then data of the monkey’s brain activity was sent in almost real time over the Internet to the CBi in Kawato’s laboratory. The CBi understood the movements of the monkey based on such data and walked in a similar way. When the movements of the CBi were displayed in front of the monkey walking on the treadmill, the monkey seemed to match its pace to that of the CBi, Nicolelis reported. The results of this experiment were viewed as a successful example of interactive BMI. (Keiichi Maeda)

Related Websites
- JST Press Release
http://www.jst.go.jp/pr/info/info461/index_e.html