Where Should I Stand? Robot Positioning in Human-Robot Conversational Groups

This paper addresses the challenge of improving robots’ social awareness within conversational groups. Many robots struggle to adapt to evolving group interactions, which can lead to negative user experiences. We introduce two innovative approaches to address this challenge: a heuristic approach, inspired by human group positioning behaviors, and a data-driven approach trained on human-human conversational group data. In our models, we include a set of features that helps robots’ positioning in conversational groups. We evaluated both approaches on the “Babble” dataset, demonstrating their reliability. The data-driven model exhibits a slight precision advantage over the heuristic model, with an average error of 9.7 cm compared to 19.4 cm.