The symbol grounding problem is about converting sensory perception into natural language. A concrete example is given in the screenshot. There is a virtual camera available which creates the tags for the mouse cursor. In the example the mouse is pointing on a yellowbox located in roomA. These tags ensure that the robot will understand a command like "Moveto roomA and grasp the yellow box" because this command is referencing to detected tags.
A semantic camera doesn't need a certain algorithm, but its mostly a GUI element. There is a rectangle on the screen and in the background the found objects are added to the status line on the bottom. All the information about walls, objects and rooms are already stored in the game engine, the camera widget formats the information only in a textual format.
No comments:
Post a Comment