British Scientists have created a robot navigation system that is based on visual recognition of a set of different objects in a particular scene,which enables the navigating robot recognize the location if it has been visited before. The robot does this by assigning a definitive description to the scene which consists of 1000 words, every two seconds. In addition, just like humans, the navigation system has the ability to combine a set of related object to each other, so that it recognizes that they might belong to the same group, such as the window, door and wheels of a car. These two principles enables the robot recognizing the previously visited locations, without getting confused if something has disappeared or something new has entered into the scene. For instance, a disappearing of a car window, car wheel and car door would only be counted as a single disappearance of a car only, not the disappearance of many objects.
Please see original article at: