A Non player character generates a quest shown on top of the screen, which gets converted into an action sequence. Each action is shown as subsentence plus a picture. For example, an "apple" is shown as a circle, while the command "turn east" is visualized as an arrow to the right.
The game has much in common with the "simon color game" except that the demonstrated pattern is more complicated.
The reason why text should be converted into pictures is because sensor data are always visual. By definition an onboard camera on a robot will perceive the environment as picture stream. Inside the stream there are no textual labels but everything is encoded in pixels. To verify if the robot has grasped an object or reached a location he needs to compare two pictures: the current situation with the goal.
No comments:
Post a Comment