Page 114 - Modern Robotics Building Versatile Macines

94 Modern Robotics

Having worked with robotic insects, rovers, and other mobile
robots, Brooks decided to create a robot that might be able to inter-
act with objects and people in a more humanlike way. Drawing on
his experience, Brooks approached intelligence, not by focusing on a
brain with sophisticated cognitive programming, but by considering
how intelligence actually arises in humans.
As Brooks noted in Flesh and Machines, “We humans are not just
products of our genes. We are also products of our social upbringing
and our interactions with the world of objects. Our culture, too, is a
product of our embodiment within the world.” This suggested that
a humanlike robot should in at least some way look like humans
and be equipped with eyes and arms that are similar to ours.
Researcher Hirokazu Kato of Waseda University in Japan had
already taken some steps toward creating a humanoid robot. His
Wabot-1, built in 1973, could walk on two legs and grasp objects
with its hands, as well as “converse” with people in a limited way.
Its 1984 successor, Wabot-2, sat at an organ, where it could read
music, play the appropriate keys, and push the pedals with its feet.
(This sort of research would culminate in the remarkably lifelike
P2, P3, and Asimo robots unveiled by Honda starting in 1998.)
Although these robots were humanoid and could even act in
humanlike ways, they did not fully satisfy Brooks’s need for a situ-
ated or embodied robot. Their behavior was essentially controlled
by a top-down central program. Harking back to Genghis and its
insect-like kin, Brooks wanted to use the behavior-based, layered
approach.
Brooks began with the eyes. Unlike a digital camera, the human
eye does not have uniform resolution and color vision across its
field of vision. The eye has a much wider panoramic view than
most cameras (up to about 160 degrees horizontally), but it has
much higher resolution and color sensitivity near its center, a
place called the fovea. When the eye sees something interesting
off to the side (for example, a running animal), it moves quickly
or “saccades” to center the vision on that spot. During this time,
a human is actually blind, though the brain creates an apparently
seamless picture.
Further research suggests that unlike the AI robot, with its inter-
nal map of the world, humans are constantly scanning, reacting,