Learning Human Actions Semantics in Virtual Reality for a Better Human-Robot Collaboration

The advent of collaborative robots has revolutionised the work concept for Small Medium Enterprises (SME), introducing quick line configuration changes and allowing a human operator to work together with a robot. Several problems are still present regarding the need to reprogram the robotic collaborator. This paper exploits the concept of Virtual Reality to instruct the robotic co-worker, allowing a non-skilled operator to teach a task effortlessly and straightforwardly. Each Skill composing the task is classified and characterised through its pre and postconditions to let the robot stop and notify the operator in case of inability to complete the scheduled action. Moreover, the system allows the operator to interact with the robot and be proactive during the task execution.