Supporting human-agent communication for explainable planning in spatial-temporal planning problems.

Area of Science:

• Artificial Intelligence
• Robotics
• Human-Computer Interaction

Background:

• Automated planning requires operators to understand and explore generated plans.
• Underwater autonomous vehicle missions present unique challenges for plan explainability.
• Existing planning models may not directly map to intuitive query concepts like distance or duration.

Purpose of the Study:

• To improve plan explainability and plan space exploration for underwater autonomous vehicle missions.
• To develop a framework that bridges the gap between planning model components and user-understandable concepts.
• To enable more effective user guidance and communication in complex mission scenarios.

Main Methods:

• Focused on the multi-agent spatial-temporal (MAST) structure as a key substructure.
• Defined model extensions incorporating concepts relevant to the MAST structure.
• Developed and evaluated new query types, including those based on numeric functions, utilizing the extended model.
• Conducted empirical and qualitative user studies in target and benchmark domains.

Main Results:

• Demonstrated the utility of the MAST structure in query formulation.
• Showcased new query types that leverage extended model concepts for better user interaction.
• Empirical studies validated the effectiveness of the new structure and query types.
• Qualitative studies confirmed improved user communication and understanding of mission objectives.

Conclusions:

• The extended MAST structure significantly enhances plan explainability for underwater autonomous vehicles.
• New query types enable users to better communicate intent and shape mission objectives.
• The approach supports more relevant information for agent responses and explanations, improving human-agent collaboration.