Sensing for Soft articulated Robots: Applications to an Out-of-the-Lab World

The present thesis considers the problem of getting robots out of our labs by focusing on novel sensing strategies for soft-articulated, anthropomorphic end-effectors. I introduce a sensing framework based on (i) the exploitation of the embedded intelligence of articulated soft robotic mechanisms, (ii) a network of off-the-shelf sensors that combine non-invasively on soft devices, and (iii) proper filtering processes, to reconstruct meaningful insight about the robot configuration, its behavior, and the unstructured environment it is interacting with. Hence, my paradigm is applied to investigate the feasibility of different tactile sensing tasks, considering tactile perception, action and interaction in out-of the lab activities. I demonstrate the application of my framework in several scenarios of practical use. The first one is the reconstruction of the posture of a soft-articulated robotic end-effector, the features of the environment it is interacting with, and the distribution of the interaction forces. Then, the paradigm is adapted to generate feedback signals that can be used to regulate the behavior of the soft end-effector during physical human-robot interaction.