Regulating Trust in Human Robot Interaction

Trust has been shown to be a crucial factor in the efficient use of automation. Trust in HRI increases the level of complexity due to the physical embodiment of the robot that shares the same space as the operator and thus resembles (partly) a "face-to-face" interaction.
Research in trust has so far been restricted to post-hoc measurements of whole interaction episodes. This is useful as an analytical tool to statically optimize the system design, but does not support trust adjustment during the interaction to ensure efficient task completion.
The goal of this project is thus: (i) identification of multi-modal trust indicators in two experimental scenarios (collaborative robotics; social robotics); (ii) data collection and training of deep learning network (LSTM) for dynamic real-time trust prediction; (iii) development of trust regulation strategies ensuring efficient use of the robot system.