Different experiments were done to build a robot that can follow a corridor. For the first version of a corridor following robot, a camera was always pointing towards the direction of the navigation, that is, towards the end of the corridor. This experiment was done with a robot called FortyTwo which has the particularity of having a turret that can turn independently of the rest of the robot.
After a change in the hardware of FortyTwo, we could not do the any other experiments with it. So another robot called Charm was used. The camera was not able to stay in a fixed direction with Charm. So we had to change completely our way of performing the corridor following.
The new algorithm done, we had problems to teach the robot how to follow corridor because of the computation time required. So we simply build another corridor following program which was not able to learn but only to follow corridors. We then used this hardwired program to teach the first program how to learn to follow a corridor.
At least, we build a robot that was able to estimate the distance that it travelled. During this experiment, the robot was also self taught. It simply went forward, took pictures of the room and recorded odometry measures during the same time.