AN ARCHITECTURE FOR OPEN-WORLD TEAM TASKS Architectural Mechanisms for Handling Human Instructions in Open-World Mixed-Initiative Team Tasks

Future envisioned mixed-initiative human-robot teams will require increasingly autonomous and capable robotic team members who can interact with their human teammates in natural ways. The challenge is to develop an integrated cognitive robotic architecture that will enable effective, natural human-robot interactions. In this paper, we present recent work on our integrated robotic DIARC architecture with its embedded Sapa Replan planner for open-world human-robot teaming tasks (i.e., tasks where not everything is known about the task objects and objectives ahead of time). We specifically focus on the required architectural mechanisms and capabilities for handling online instructions with unspecified objects and describe the novel mechanisms implemented in DIARC and Sapa Replan that enable online open-world tasking.