We have long trusted the effectiveness of robot autonomy. However, we have gradually come to recognize that merely pursuing autonomy might not create useful robots the way we had expected. In 2006, MotionLab recognized the crucial role of spatial-understanding in robotic technology.
Here a one-year-old boy, John, is playing alone in the living room. He suddenly feels lonely and wants to find his mother. He crawls along a hallway with an expectation of finding her in the kitchen. He acts in this way even though she is NOT directly visible. His action proves that he grasps the spatial relations, a virtual map, of the house in his brain, even though no one has taught him the house's layout. Further, he solves the shortest-path-finding algorithm using this virtual map, and executes the computed path. John deeply understands his space.
John also starts to associate names with places: associating the name of "kitchen," "bathroom," "bed room," and "garage" to corresponding places so that he can communicate with his parents intelligently. The use of names is also a part of his spatial-understanding capacity.
The next question is how a robot can attain this super-intelligent capacity.