A robot from Brown University called Baxter has learned to perform a task from a different robot called PR2 from Cornell University through a central repository called RoboBrain. PR2 was taught to perform simple demonstration tasks which was then stored in RoboBrain's knowledge database. From there, Baxter took what PR2 learned and was able to perform the task in a different setting and situation. Assistant professor at Brown University Stefanie Tellex said, “It’s pointing in an interesting direction. When you put a robot in a new situation—and in the real world it happens in every room the robot goes into—you somehow want that same robot to engage in autonomous behaviors.”
Tellex says, “this is really a baby step toward that vision. There are a lot of remaining technical challenges.” The challenge with transferring learning between different robots is that both robots are likely built very differently. Low-level commands may turn out very differently when joints are positioned differently. In the future, scientists aim to be able to have the robots translate information in relation to its physical body relative to other robots. This will involve developing a way to have commands be transferred between two different platforms. With increased bandwidth and cloud computing capacity, it's becoming more and more feasible for robots to share knowledge and learn from one another.