Visual Memory

The Visual Memory stores landmarks and detected obstacles, with their location constantly updated using odometric information. To deal with the imprecision in odometry we use, again, a fuzzy approach. The odometric information coming from the robot is indeed fuzzy information about its motion, used to recompute the location of the objects stored in the Visual Memory. The imprecision of this motion is higher when the robot turns, and lower if it moves straight.

As the robot moves, the imprecision on these locations grows unless the landmarks are recognized again by the Vision system (which obviously reduces their location's imprecision). When the imprecision about the location of a landmark reaches a given upper threshold, the landmark is removed from the Visual Memory. The idea behind this being that the Visual Memory only remembers those landmarks whose location is precise enough.

The information stored in the Visual Memory is treated by the Navigation system in the same way as the information coming from the Vision system. The only difference is that the Visual Memory will be more imprecise than the Vision system. The Pilot system also uses this information to avoid colliding with remembered obstacles and landmarks.