4. Domain/Environment Engineering Issues
Keeping our environment simple was actually a complicated feat. We wanted to make sure the images we got from the camera were binary images containing only one solid line that would either be straight or would branch into two lines (for fork). These are the variables we took into consideration to control the output of the images from the web cam - high contrast between floor and line of tape, lighting, tank turning error, Z80 processor transmission speed, web camera reliability.
4.1 High Contrast
To save on processing during the reduction step, we wanted to avoid "blob" calculation because that would mean we have to make two costly scans through the image - one scan to mark the blobs, one to keep the largest blob. To do this, we made sure that we chose contrasting colors for the background and line. During the experiment, we used a dark blue carpet with bright yellow tape. The reduction from color to binary images was trivial at that point (see 'Color Chooser' and 'Data Reduction' in previous section).
The lines on the floor must be evenly lit, so the lights must be adjusted so that they are not too bright or too dark. Consider a pitch black room - all you see is background. Consider an extremely bright room - 80% of the image would be considered "line." Low and even lighting seem provided the best results for our experiment.
4.3 Tank Turning Error
An important factor in making decisions was the minimization of error in turning the tank. The Sentinel had a minimal turning error of about + 5 degrees, and this made it difficult for the web cam to provide us with an image of a "straight line" (defined as top center to the bottom center of the image).
4.4 Z80 Transmission Speed
The Z80 board sent the commands to the tank at a very slow pace. Although images and decisions were processed very quickly, the response time of the Z80 board was not as good. This intermittent transmission resulted in a non-continuous movement of the tank. The tank only processed approximately two commands every second.
4.5 Web Camera Reliability
The wireless web cam posed issues with reception. Whenever the image started to "break" up, the image distorted as a result of scan lines, fuzz, and static. As these mal-formed images were processed through the decision algorithm, incorrect decisions were made. To minimize this error, we attached the infrared receiver to the ceiling, maximizing the range of reception (this was placed above the center of the track.)