The Real Scenarios

The different classes of scenarios on which the experimentation has been carried out are the following:

1. Single landmark: in this class of scenario there is only one landmark, which is the target, and no obstacles. This class of scenarios is used to check that the robot is able to reach a target when there are no references to it and there exists a clear path to the target.

2. Single landmark and obstacles: these scenarios are composed of a single landmark which is the target, and several small obstacles that do not occlude the target, but force the robot to avoid them in order to get to the target.

3. Several landmarks: in these scenarios there are several landmarks, one of them being the target, but no obstacles (apart from the landmarks themselves, which are obviously seen as obstacles). In these scenarios the Navigation system is able to build a map of the environment, and we will check how good it is.

4. Several landmarks and obstacles: in these scenarios we add obstacles between the landmarks of the previous scenarios so that they block the robot and it is forced to compute diverting targets to reach the original one. In these scenarios the Navigation system is also able to build a map of the environment, including the detected blocking obstacles.

Some pictures of the different scenarios can be seen in Figure 6.7.

The first two classes of scenarios are very simple, and the experiments on such scenarios just check the very basic behavior of reaching a target through a quite clear path. In these scenarios the target is visible all the time, as the only obstacles are small ones, therefore not occluding the view field of the camera. The real tests are in classes 3 and 4, as the target may be occluded by other landmarks, and the path to the target might be blocked by landmarks and obstacles. Thus, in these scenarios, the robot must make use of its navigational skills.

We impose the restriction of the objects on the environment (that is, landmarks and obstacles) be static, so their location cannot change during a trial. If that were allowed, the computed relation among landmarks would be inconsistent, and thus the $\beta$-vector computation would not be valid at all.

Figure 6.7: Top left: one of the obstacles used in the environments. Top right: scenario 1. Middle left: scenario 2. Middle right: scenario 3. Bottom left and right: scenario 4.