Generation of Reaching Motions for Flat Cable Insertion Task Using Simulation Learning and Domain Adaptation for Industrial Robots

• DOI: 10.20965/ijat.2025.p0280
• Publication Date: 2025-05-04T15:02:06Z
• Authors: Yuki Yamaguchi; Shinsuke Nakashima; Hiroki Murakami; Tetsushi Nakai; Qi An; Atsushi Yamashita
• Keywords: Domain Adaptation

In this paper, we propose a method for generating reaching motions the insertion of flat cables. Despite need to insert cables into sockets in circuit assembly various electronic devices, there has been little research on automating that are already fixed one side board. regard, focus generation posture grasping such Our uses deep reinforcement learning simulation environment, and features extracted from image pose manipulator used as states. For transfer environment real-world use CycleGAN-based domain adaptation method. We conducted experiments under several different conditions verify operation trained agent. The results demonstrated success rate generated exceeded 70% all conditions.