SwapBot: Tool-Centric Robot Manipulation via Interchangeable End-Effectors and Tactile Sensing

TL;DR

SwapBot introduces a modular, tool-centric robotic manipulation system that uses low-cost, interchangeable adapters to convert household tools into robot end effectors. By integrating tactile sensing and enabling fast autonomous swapping, SwapBot supports diverse contact-rich tasks like cooking and stirring, with imitation learning policies trained on tactile-augmented demonstrations showing 15–25% higher success rates than RGB-only approaches.

Why It Matters

This work addresses a critical gap in real-world robot manipulation by enabling versatile, task-specific tool use without requiring custom grippers. The modular design lowers cost and increases adaptability across different robots and environments, making advanced manipulation more accessible for household and service robotics.

Key Findings

• Tactile-augmented demonstrations improve imitation learning success rates by 15–25% in contact-rich tasks.
• Modular adapters enable fast, autonomous tool swapping with diverse tools, including articulated ones.
• User studies show reduced task completion time and operator workload during teleoperation.