Leonel Rozo and Sylvain Calinon, have programmed a robot to help them build an IKEA table, using a process called kinesthetic teaching. The robot used a force sensor and a vision tracking system to learn the movements a researcher was taking it through. The user grasps the robot by hand to demonstrate how it should collaborate, and the sensor - mounted at the wrist of the robot and the vision tracking system was used to record the position and orientation of the table legs that need to be mounted at four different points on the table top. The robot 'learns to be compliant' to let the user re-orient the table top in a comfortable pose to screw the corresponding table leg. Once the user starts to screw the leg, the robot becomes stiff to help.
The robot assists the IKEA builder in flipping the tabletop over and becomes stiff while the builder screws the legs of the table in. The behaviour was not pre-programmed, but is instead learn by the robot by extracting the regularities of the task from multiple demonstrations. To be a real help the Robot would need to translate the instructions into something usuable, be able to identify which is the correct screw to go where. Otherwise you will end up hitting the screws in with a hammer like normal.
The robot was not designed to make furniture. It is part of a more important AI project which is focused on transferring movements from humans to robots. The researchers said that there is a need is arising for robots that do not just replicate the task on their own, but that also interact with humans in a safe and natural way to accomplish tasks cooperatively.
'Robots with variable impedance capabilities opens the door to new challenging applications, where the learning algorithms must be extended by encapsulating force and vision information,' the report said.