The first corridor follower built using the robot FortyTwo was not perfect. This was due to the way the robot was trained. But the results let us hope that it is possible to build a corridor follower using FortyTwo. We simply could not test this corridor follower because of the change in the hardware.
The version of the corridor following robot using automatic learning gave good results. The robot was trained in one corridor and tested in three different corridors and it worked successfully. The program designed to teach used ceiling as landmarks, but the program that learns used lights as well as ceiling to do the same task. This brought robustness to the program. During the tests, the travelled paths were very close to their mean path. The original orientation of the robot for successed taken paths was also studied. It reveals that the algorithm is relatively robust to the choice of this angle provided that it is in a fixed range (from -10 degres to 30 degres).
Finally, the distance estimation experiments showed that for long distances it performs better than the internal odometry based on wheels. For shorter distances the internal odometry performs better.