Robotic Assembly of Deformable Linear Objects via Curriculum Reinforcement Learning

The automated assembly of flexible objects presents significant challenges. Although significant progress has been made in the assembly of rigid objects, the methods used for rigid objects cannot be directly applied to flexible objects due to their infinite degrees of freedom. This study proposes a reinforcement learning (RL) based method for deformable cable insertion tasks executed with a universal 2-finger gripper. Firstly, a vision-based detection method is employed to monitor the cable's state in real time, while a state classifier is introduced to provide real-time reward feedback for RL training. Secondly, an adaptive curriculum learning (CL) method is proposed to adjust the initial degree of cable bending through the success rate in the training process, allowing the RL agent to learn progressively from easier to more difficult tasks. The validation experiments were conducted on a type-C cable insertion task, where the robot grips the cable portion of the electrical connector. The results indicate that our method is capable of adapting to various degrees of cable bending, successfully handling cable configurations bent up to a maximum of 40° from its straight, unbent state, with an assembly success rate of over 90%.