Robotics Reference
In-Depth Information
Cohen's ambitious goal when he started was to teach the computer
the rules of artistic composition on which his own work was based. In-
stead his project has turned into the world's most advanced computer
artist. In 1980 Cohen began to examine the scribbling behaviour of
young children, focussing on the moment in the scribbling process at
which the first part of a scribble migrates outwards and becomes an en-
closing form for the rest of the scribble. Cohen perceived this to be the
moment at which the child becomes aware that the scribble means some-
thing, and he found the geometry of enclosure, the physical relationship
of the enclosing form to what is being enclosed, quite baffling. Cohen
attempted without success to simulate this early human drawing, but
he became convinced that the range of forms AARON could generate
would be greatly enhanced if its steering strategy could be made to find
its way around a pre-existing “core figure”, the computer equivalent of
the way that a child's initial scribble partly determines the path it later
traces to enclose it. This conviction proved, in due course, to be justified.
The construction of simple core figures, plus a simple strategy for trac-
ing a path around them, yielded forms of sufficient complexity to enrich
AARON's drawings. For example, AARON employs a sort of internal
scribble when it draws trees. From the knowledge Cohen has given it
about how trees grow, AARON grows its own tree skeleton, and then
draws a line around it to create the complete image of the tree.
AARON's approach to the creation of drawings is based on the idea of
beginning with something simple and building on it. AARON generates
a drawing by using its knowledge about the objects that it draws and
its knowledge about how to build visual representations of those objects.
These two types of knowledge interact with each other continuously in
complex ways and are fundamental to AARON's drawing expertise. All
of its knowledge has been given to it by Cohen, but it is AARON that
decides how to use this knowledge.
Some of AARON's knowledge is easy to represent, for example, how
long are a person's arms and legs. Cohen simply make a list of parts, such
as the left-upper-arm, the torso, etc. Each part is represented by a list
of all the data points in that part, giving the position of each point in
relation to the “origin” (or end-point) of the part. Thus, the origin of
“left-upper-arm” is “left-shoulder”, while the position of “left-elbow” is
specified to be at some position relative to “left-upper-arm”, such as two
inches below, one inch to the left, three inches in front. And AARON
knows that the “left-upper-arm” is attached to the torso, but it can only
