Page 153 - Modern Robotics Building Versatile Macines

RADICAL ROBOTICIST 133

table with bicycle wheels. The robot
sent images from a television cam-
era back to its controlling com-
puter. Mobile robotics had become
an exciting and promising field,
as shown by the funding of Cart
research by the Defense Advanced
Research Projects Agency (DARPA),
the National Science Foundation,
and the National Aeronautics and
Space Administration.
The Cart’s control program
had several subroutines. One,
called the interest operator, tried
to identify “regions of interest”
The Stanford Cart, photographed
in one of a series of photographs
some time during the 1970s. The
taken by the Cart’s camera. (For rather humble-looking robot was
example, these might be places the test bed for pioneering work
where there were edges indicat- in mobile robot navigation. (Photo
ing objects and thus potential courtesy of Rodney Brooks)
obstacles.) The correlator rou-
tine then looked for matching
features in another picture. The
camera-solver routine then triangulated the shifted positions
between the two pictures in order to determine the distance to
the objects of interest. (This is a process that human eyes can
perform in a fraction of a second.)
Once the obstacles had been identified and located, the navigator
routine planned a path to the destination that avoided the obstacles.
The path was then translated into driving instructions that moved
the Cart about three feet (1 m) along the path. Another set of pic-
tures was then taken, and the process would be repeated until the
Cart reached its destination. The relative slowness of the available
minicomputers at the time meant that the Cart drove in a jerky
fashion, with 10 to 15 minutes between movements.
The Cart was the first autonomous mobile robot that could plot
paths around obstacles, but if one could see the world as the robot
saw it, there would be no objects as such, or even wireframe outlines.