Page 255 - Introduction to AI Robotics

238
6 Common Sensing Techniques for Reactive Robots

a.

b.

c.

Figure 6.26 Situations and resulting images of a.) flat surface, b.) an obstacle, and
c.) a negative obstacle.

duce a fairly coarse pattern; they don’t project a line or grid onto every pixel
in the image. This means the device is overall less demanding, and therefore
less expensive to build.
Light stripers were popular in the late 1980’s and early 1990’s, and labora-
tories such as the University of Pennsylvania’s GRASP Lab under the direc-
tion of Ruzena Bajcsy produced excellent results in extracting 3D informa-
tion. However, these efforts focused on using the depth map to recognize an
object under laboratory conditions. The results did not transfer particularly
well to mobile robots. Reactive robots are not concerned with recognition,
so many of the algorithms were not particularly useful or provided a quick,
reflexive response. Also, in the open world, objects were often the same color
as the projected light or near enough to it to confuse the vision system. The
amount and direction of lighting could also confuse the striper, with brightly
lit rooms making it difficult to see the bright laser or structured light.