Monitoring and Guiding User Attention and Intention in Human-Robot Interaction

Abstract

Abstract—A robot interacting with humans and attempting to generate effective social interaction and intervention behaviors benefits greatly from being able to understand and predict the underlying intentions of actions in context. Related work on collaborative discourse suggests that intention can be described in terms of either goal-directed task completion or communicative behavior directed to other collaboration partners. This paper describes early work on a generalizable framework for estimating the attentional space of a human interaction partner, providing context for grounding action in terms of intentions, and using this model to perform contextualized robotic intervention and ambiguity resolution. We describe an experiment aimed at applying and validating the framework in a simple collaborative human-robot interaction scenario. O I.