HOTSPOT: An ad hoc teamwork platform for mixed human-robot teams

Area of Science:

• Robotics
• Artificial Intelligence
• Multi-Agent Systems

Background:

• Ad hoc teamwork research traditionally focuses on agent-only systems.
• Collaboration between unknown agents without prior communication is a key challenge.
• The potential of ad hoc teamwork in human-robot interaction remains largely unexplored.

Purpose of the Study:

• To propose HOTSPOT, the first framework for ad hoc teamwork specifically designed for human-robot teams.
• To address the challenges of integrating robots as ad hoc agents collaborating with human teammates.
• To facilitate effective decision-making and communication in mixed human-robot teams.

Main Methods:

• Developed a decision-theoretic module for task identification, teammate identification, and planning.
• Integrated a natural language processing-based communication module to interpret human-robot interactions.
• Evaluated the framework through a cooperative object collection task involving a mobile robot and a human.

Main Results:

• The HOTSPOT framework successfully enables a robot to act as an ad hoc agent with human teammates.
• The decision-theoretic and communication modules effectively manage task execution and interaction.
• Demonstrated the framework's capabilities in a realistic human-robot collaborative task.

Conclusions:

• HOTSPOT represents a significant advancement in ad hoc teamwork for human-robot teams.
• The proposed framework has broad applications in scenarios requiring seamless human-robot collaboration.
• Future work can expand upon the decision-making and communication capabilities for more complex tasks.