Artificial Intelligence and Computer Vision Techniques for Human-Robot Interaction

The aim of this thesis is to investigate techniques to improve the interaction between humans and collaborative robots (i.e. robots which shares the workplace with humans without barrier and without exposing the human to risks) exploiting Artificial Intelligence and Computer Vision techniques. To highlight the effectiveness and the potential of Computer Vision (CV) applied to Robotics and to analyze how the Human-Robot Interaction might be powered leveraging the CV exper- tise, three cases of study, set in three different contexts, are developed for the UR5 cobot (developed by Universal Robots) and the required algorithms are developed using state-of-the- art methods. The first is set in a television studio where the television presenter moves in the studio and the robotic arm follows him using a face-tracking algorithm. The second is set in an industrial context where an operator has to process workpieces which are numbered from 1 to 5. The operator is able to select the workpiece to be processed using gestures, and the camera-equipped robot responds accordingly. The third is set in a hospital: a patient is in its bed and, unable to move, he can communicate with the robot with gestures and ask for drugs, water or telephone. These cases of study are simulated in the Polyscope simulation environment, the propri- etary software of Universal Robots, and RoboDK which is a universal environment for robot simulation. The computer code required for CV and AI tasks, which is developed in Python using the OpenCV library, is thoroughly analyzed in this work, as well as the state-of-the art models that were exploited and the related theory.