Human-inspired motion primitives for a safe and effective HRI and autonomous manipulation

Recent advancements in robotics are pushing toward a more strict interaction between machines and their surrounding environment, including humans, in the general framework of co-botics, in both industrial and assistive applications. This comes with the need of guaranteeing i) the acceptability of the technology and ii) the users’ safety. One of the key characteristics that a co-bot should have to satisfy these requirements is to behave as a human would, especially for what concerns the predictability of the robot movements. The intent of our work is to develop a framework for the generation of human-inspired movements for a generic redundant manipulator that works in an unstructured environment. This manuscript reports on recent advancements of our contribution to this line of research, with a special focus on the identification of a basis of motion primitives, which can be used as key ingredients for the planning of human-like motions. Then, we discuss how the feed-forward implementation of these human-like motions can be integrated with a tactile-based feedback policy, when a purposeful interaction with an object happens. These results could open also interesting perspectives for autonomous robotic manipulation.