Skills learning in robots by interaction with users and environment

The fast technological evolution and dissemination of multimodal sensors and compliant actuators bring a new human-centric perspective to robotics. The variety of human-robot interactions that stem from these new capabilities unveil compelling challenges for machine learning. The aim of this paper is to provide robots with a representation of rich motor skills able to handle recognition, prediction, synthesis and refinement in a continuous and synergistic way. It also requires to be robust to various sources of perturbation, persistently arising from the environment, from the user, and from the robot. One important challenge in this direction is to devise an encoding scheme that is able to generalize tasks to new situations, that can potentially act in multiple coordinate systems, and that can exploit the modern compliant control capabilities of robots to generate natural, efficient and safe movements for the surrounding users.