The Task-State Coordination Pattern, with applications in Human-Robot-Interaction

Abstract

We consider interaction a powerful enabling technology for robots in human 
environments. Besides taking commands or reporting, many other uses, 
such as interactive learning, are already being explored. However, HRI also 
poses systems engineering challenges that may hinder its adoption. To address these, we advocate a general coordination pattern for task execution: The Task-State Pattern. Crucially, it separates interaction coordination from task-level control, thus enabling independent, but integrated, development.

In the pattern, tasks are represented using both a general, re-usable task coordination model and a task-type dependent specification. We have introduced a coordination model rich enough to support a powerful user experience, but still general enough to accomodate a variety of tasks, thus simplifying architecture and integration. Furthermore, because it is re-used
in many places, it provides an attractive target for tool support.