Vision System

The Vision system is able to identify new landmarks in the vision field of the camera and is also able to recognize previously identified landmarks. This system does not bid for any of the available services. Again, although this system is not on the focus of the thesis, we have had to develop a simple Vision system for carrying out the experiments. A detailed description of the vision system developed to recognize indoor landmarks is given in Chapter 6.

The Vision system is simple but robust enough to correctly identify the landmarks. Thus, there is no uncertainty about the presence of a given landmark. However, there is some imprecision about its location, since the Vision system only gives approximate distance and angular information. To deal with this imprecision we use the fuzzy techniques described in Section 3.2.

The goal of this thesis is to develop a vision-based navigation system that does not use any specialized localization device (e.g. GPS) nor odometric information. However, we found that it was very restricting for the Navigation system to use only the visual information available after processing each viewframe. Firstly, because it is very difficult to have more than three landmarks on the view field, since it is very narrow, and the beta-coefficient system needs to have at least four visible landmarks in order to create a new $\beta$-unit. But even if four landmarks were in the view field, they would probably be highly collinear, which is not a good configuration for creating $\beta$-units. Secondly, it was a very unrealistic behavior to completely forget the landmarks that were not in the view field, even though they had been recently seen. We thought that adding the ability of remembering what has been previously seen would improve the behavior of the robot. Moreover, as it has already been mentioned, we want the robot to imitate the navigational behavior of humans and other animals, and we certainly have the ability of remembering what has been recently seen. A short term memory, called Visual Memory, implements this ability, and it is part of the Vision system.