Spatial Formation Model for Initiating Conversation

In a situation where a robot initiates conversation with a person, when is the appropriate timing and where is an appropriate position from which to say the first greeting word? In this study, we analyze human interaction and establish a model for a natural way of initiating conversation. The model mainly concerns the participation state and spatial formation. When a person is going to participate in a conversation, at a moment when a particular spatial formation occurs, she would feel that she is participating in the conversation; once she perceived her participation she would try to maintain particular spatial formations. There are theories in human communication for these concepts, but they only cover the situation after people have started to talk. We build a model that precisely describes the constraints and expected behaviors for the phase of initiating conversation. The proposed model is implemented in a humanoid robot, and it is confirmed as effective in an evaluation experiment based on a shopkeeper scenario.