Researchers control robot with monkey's brain signals

by Martyn Williams

January 15, 2008 —

Scientists in Japan have succeeded in controlling a humanoid robot with signals
picked up in the U.S. from a monkey's brain and transmitted across the Internet,
they said Tuesday.

The research, which represents a world's first according to the Japan Science
and Technology Agency (JST), could be a first step toward giving doctors the
ability to restore motor functions in severely paralyzed patients. It can also
contribute to the development of robots that move more like humans, JST said
in a statement Tuesday.

In the tests, scientists led by Miguel Nicolelis at Duke University in North
Carolina trained two monkeys to walk on their legs on a treadmill. The activity
of neurons in the leg area of the monkey's brain was recorded while the monkey
walked and decoded into predictions of the position of their leg joints.

These predictions were then sent across the Internet to Kyoto where they were
used to control a robot. A live video signal of the robot was relayed back to
the monkey to provide feedback.

The robot, called CBi for Computational Brain interface, is about the same
size as a human at 155 centimeters tall and weighs 85 kilograms. It has 51 degrees
of freedom of motion and was developed by JST and Christopher Atkeson of Carnegie
Mellon University's Robotics Institute to enable such neuroscience research.
The hardware side of the robot was developed by Sarcos, a Salt Lake City robotics
company.

The results of the work are groundbreaking, according to JST, although much
remains to be done before it can be worked into something useful. As part of
the ongoing research, the teams are looking at sending back more complex feedback
to the brains of the monkeys.

In recent years, robotics researchers have been increasingly studying how to
make the movements of robots more lifelike. Robots like Asimo, developed by
car-maker Honda, are being positioned as future companion robots that could
either work alongside humans or carry out tasks for them. One of the many issues
that needs to be tackled before such a dream can be realized is increasing the
mechanical complexity of the robot while simultaneously developing more advanced
control systems.

IDG News Service