This research introduces a novel approach for enabling robots to intelligently manipulate objects using tools without grasping them, integrating language model planning with physics-based controls.

• Combines LLM-based high-level planning with specialized low-level controls for precise manipulation
• Introduces a manoeuvrability score to help robots choose optimal tool-object interactions
• Demonstrates success across various challenging manipulation tasks in both simulation and real-world settings
• Shows potential for robots to reason about tool use similar to biological intelligence

This breakthrough in engineering allows robots to achieve human-like dexterity in complex manipulation tasks, potentially revolutionizing applications in manufacturing, service robotics, and assistive technologies.

Non-Prehensile Tool-Object Manipulation by Integrating LLM-Based Planning and Manoeuvrability-Driven Controls