Robotic etiquette : structured interaction in humans and robots

Visual nonverbal behaviour plays a significant role in allowing humans to structure their interactions meaningfully, thus facilitating communication between individuals. Such behaviour may be exploited in the design of socially intelligent robots. While much visual behaviour (e.g. facial expression) is hard to detect using a machine vision system, especially in unconstrained situations such as natural interaction, simple gestures and the positions of interactants are relatively easy to detect. This paper begins with a description of some related work, followed by an overview of some systems that track and/or interpret human movement. This is followed by a discussion of what is known about the structure of human interactions and a detailed example of a particular kind of structured interaction (the greeting). Ways in which interactional structure can be exploited in the development of socially intelligent robots are considered: finally, a specific application involving the use of robots in the rehabilitation of autistic children is described. We refer to this approach, using knowledge of interactional structure both to generate social behaviour for a robot and to interpret interactions for a machine vision system, as interactive vision. The work described in this paper is still in its early stages and there are no experimental results to report as yet: the purpose of this paper is rather to outline ways in which sociological and psychological work on human interactional behaviour can be exploited in socially intelligent agents and machine vision systems.