Enhancing Human-Robot Interaction by Detecting and Modulating Information Flows

Communication, the flow of information between agents, is vital socially acceptably robot behaviours. Understanding and utilising social information is essential for achieving such behaviours. This research investigates the detection, analysis, and application of social information flows through the lens of information theory. This research comprises three stages: detecting and analysing social cues, applying transfer entropy to enhance human-robot interaction (HRI), and exploring real-world applications. We have proposed a framework for social cue detection and analysis, demonstrated across three human interaction settings: person-following, object-handover, and group-joining. This framework provides a systematic workflow that yields reliable results. In the second stage, our simulations and human studies have shown transfer entropy's effectiveness in improving social communication within a reinforcement learning context. In the third stage, we aim to validate our framework through practical user studies, enhancing its adaptability and exploring influence modulation across diverse HRI scenarios.