Randomized trial demonstrates the effectiveness of a hardware-independent skill library for robots, highlighting software reusability.
Key Points
This research aims to develop a reusable control program for robots that utilizes a hardware-independent design for manipulation skills.
Proposed a Learning-from-Observation (LfO) framework with a pre-designed skill library.
Defined a skill-agent set that covers all possible actions for robot manipulation.
Demonstrated the execution of skill agents on two different robots and end-effectors.
Successfully executed the same task model representations on Nextage and Fetch robots with different end-effectors (Shadow Hand-Lite and parallel gripper).
Demonstrated the practicality of the skill agents in facilitating software reuse across various robots.
Defined necessary skill agents to minimize development efforts for new robot hardware.