This video shows how a humanoid robot can learn incrementally new skills by interacting with a human instructor. By using active teaching methods that puts the human teacher "in the loop" of the robot's learning, this video shows that a skill can be efficiently transferred by interacting socially with the robot. In a first phase, the robot observes the user demonstrating how to lift a foam dice while wearing motion sensors. The motion of his/her two
arms and head are recorded by the robot and encoded probabilistically in a Gaussian Mixture Model. The robot then tries to reproduce the skill while the user observes the imitation attempt. Due to the different embodiment between the robot and the user, the robot first fails at reproducing correctly the task. In a second phase, the user helps the robot refine its skill by kinesthetic teaching, i.e. by grabbing and moving its arms throughout the task to provide the appropriate scaffolds. The model of the skill is then updated by an incremental Expectation-Maximization learning algorithm. After three demonstrations, the robot finally reproduces perfectly the skill on its own.
- YT Poster.