Automatic Cable Routing in 3D Electrical Panels Using Deep Reinforcement Learning

The advancement of modern industries and electronic devices has made cable routing in three-dimensional (3D) environments an essential design process across various fields, including machinery, automobiles, aircraft, and electronics. Traditional cable routing is often manually performed by experts, leading to variability in quality and potential errors depending on the operator's skill. Moreover, frequent design changes can significantly increase time and costs. To address these issues, this study proposes an automatic cable routing method for 3D electrical panel design using deep reinforcement learning. The approach incorporates a sequential task generation process for path exploration and employs a curriculum learning framework to train routing agents effectively. The trained agents generate natural 3D cable models based on the identified paths. Experimental results demonstrate that the proposed method significantly improves path planning efficiency while satisfying design constraints. This study highlights the potential of combining grid structures and reinforcement learning to solve cable routing problems and improve practical applicability in panel design.