A Model of Continuous Intention Grounding for HRI

We present a formal model of intention grounding for robots in HRI which models the user and robot’s own current intentions simultaneously. Central to the model are interactive state machines, and strength-of-evidence functions, which are evaluations of the degree to which an agent’s current intention is being communicated publicly. The model has so far been used in a simple pick-and-place robot with a speech interface, but we argue this can be generalized to other robots such as semi-autonomous vehicles, providing potential for automatic evaluation and active learning.