What We Can Learn From Robots
From: Technology Review - 01/2005 - Vol. 108, No. 1, P. 54
By: Gregory T. Huang

Director of Japan's ATR Computational Neuroscience Laboratories Mitsuo Kawato
values robots for the insight they can provide into human brain functions, a
concept that deviates from the usual economic or assisted-living motivations
behind robot development, according to Carnegie Mellon robotics expert
Christopher Atkeson. Kawato is convinced that experiments with humanoid
robots can yield streamlined simulations of brain-cell behavior that can be
compared to the actual workings of neurons in human and primate brains using
sophisticated imaging methods. The data from such research could be applied
to the creation of therapies for brain damage as well as neurological,
cognitive, and behavioral disorders. ATR researchers are employing humanoid
robots such as Dynamic Brain (DB) to test neuroscience theories: Kawato and
fellow ATR scientist Gordon Cheng believe people use "internal models" to
measure connections between neural signals and subsequent body movements, and
have extended that hypothesis to DB by having the robot use software to
compute what commands will generate the proper series of motions needed to
fulfill a certain objective. Determining how big a role a brain's operations
play in a robot's execution of tasks is the focus of a project whereby human
subjects learning to use an unfamiliar tool are analyzed via magnetic
resonance imaging in the hopes that the acquired knowledge will lead to
better robots. One of the hoped-for goals of such research is a remote
brain-machine interface that will enable the user to participate in
geographically distant events. Another objective is to make robots more
autonomous, and an $8 million upgrade to bring DB's anatomy, neural
architecture, power requirements, and strength to a more human level will be
employed to study gait disorders and falls among the elderly. Kawato is also
urging Japan's government to help fund a global initiative to build a robot
that matches a five-year-old child in terms of cognitive and physical
ability.

Links:
Kawato Dynamic Brain
http://www.jst.go.jp/erato/project/kgd_P/kgd_P.html

Kawato Dynamic Brain Project
http://www.kawato.jst.go.jp/

