Statistics Since Feb 24, 2011:
Download flv 15M avi 21M wmw 34M |
|
|
| This demo shows an attempt on incremental robot learning from demonstration. Based on previously learnt knowledge about a task in simpler situations, a robot learns to fulfill the same task properly in a more complicated situation through analyzing comparative demonstrations and extracting new knowledge, especially the constraints that the task in the new situation imposes on the robot's behaviors.
In the experiment, the teacher demonstrates a positive and a negative example, where the task is the same, to pick up the can on the inside end of the board. In the negative example, the teacher picks up the inside can directly, causing the outside one to fall; in the positive example, the teacher puts the outside can on the table first, then to pick up the inside one. The robot can generate these two sequences with the current KB, but will always choose the wrong one because it is shorter and the current KB does not contain knowledge about this task situation---there are no rules predicting the falling of items or constraints prohibiting falling of items. So this is a substantially new situation to the robot. The experiment shows that, through the learning based on the previous knowledge, the robot can complete the task while avoiding items falling. |
Statistics Since Feb 24, 2011: |