Combining synchrony and shape detection to sustain the robot focus of attention on a selected human partner

Abstract

The present study deals with the problematic of attentional mechanism allowing to initiate and to maintain Human Robot Interactions (HRI) by orienting the robot's visual focus on interacting human partners. In our previous work, we took inspiration from human psychological and neurological data which suggest that synchrony is an important parameter for human-human interaction. We proposed synchrony as a way of interacting and presented a synchrony-based architecture capable of selecting the human partner and of locating the focus of attention. To deal with the problematic of initiating the HRI, we proposed, in our recent works, a neural model permitting to focus the robot visual attention on a selected partner on the basis of synchrony detection between its own dynamics and the human movements. This model maintain the interaction and the robot's focus of attention while the partner moves in synchrony. Consequently, the interaction is interrupted if the partner stops moving. For a more realist HRI, the agents have to be able to switch their roles (turn tacking), as a result, they could alternate moving and static interaction phases. In this case, we propose here to complete the previous neural model by adding a shape based attentional mechanism. After initiating the interaction on the basis of synchrony, the robot will automatically learn to recognize the selected partner and maintain its attention with the human during unsynchronized phases of interaction.