Robotic Disassembly Task Training and Skill Transfer Using Reinforcement Learning

Abstract

This paper proposes a platform for robots to learn disassembly tasks based on reinforcement learning (RL) techniques. The platform is demonstrated by a robot learning the skill of removing a bolt along a door-chain groove in a data-driven way, where the clearance between the bolt and the groove is less than 1mm. Furthermore, the relationship between the performance of the learned skills and the precision of the robot is studied with a method to control the robot's precision by adding uncorrelated zero-mean Gaussian noise to the robot's actions. Finally, the transferability of the learned skills among robots with different precisions is empirically studied. It has been found that skills learned by a low-precision robot can perform better on a robot with higher precision, and skills learned by a high-precision robot have worse performance on robots with lower precision.