Interactive Adjustment of Agent-Based Cooperative Tasks Driven by Multi-View Collaboration

Area of Science:

• Artificial Intelligence
• Human-Computer Interaction
• Robotics

Background:

• Flexible and complex task execution in agent-based systems requires interactive adjustment.
• Current methods lack intuitive visual interaction, hindering real-time human intervention.

Purpose of the Study:

• To propose an interactive adjustment framework for agent-based cooperative tasks using multi-view collaboration.
• To enhance user interaction and task flexibility in complex scenarios.

Main Methods:

• Developed a Dual-Structured Graph for unified task and agent process visualization.
• Integrated a large language model (LLM) for interpreting diverse user commands.
• Implemented an integrated platform with three views for multi-view collaboration.

Main Results:

• The Dual-Structured Graph provides a unified task representation.
• LLM-driven interaction enables accurate command interpretation and execution.
• The multi-view platform supports intuitive and flexible task adjustment.

Conclusions:

• The proposed framework and platform effectively address limitations in current interactive adjustment approaches.
• Multi-view collaboration enhances human intervention in agent-based cooperative tasks.
• User study validates the framework's effectiveness in real-world scenarios.