Robot learns to insert bending rods precisely with world model guidance
RodForesight: A World Model Enhanced Diffusion Policy for Slender and Material Agnostic Rod Insertion
Robotics
Summary
Inserting thin, bendy rods into tight spaces is hard because the rod's tip can move unpredictably when it bends. This paper presents RodForesight, a method that first moves the rod close to the hole using cameras and then carefully uses a special AI approach to predict and correct rod alignment. This method tries out multiple possible moves before picking the best one, leading to more successful insertions than earlier techniques. The authors show that this approach improves success rates on the task by combining vision, prediction, and decision-making.
What this means in practice
- •For industrial automation teams: Improve robotic assembly lines by enabling precise insertion of thin, flexible rods with varied materials and bending behaviors.$Commercial implications: Enables manufacturing robots to handle delicate rod insertions more reliably, reducing defects and downtime.
- •For robotics system integrators: Use a two-stage approach combining visual guidance and predictive modeling to enhance robotic tasks involving objects that deform during manipulation.
Authors
Chuanbo Yu, Mingyu Yue, Yan Lyu, Chuhan Song, Peng Wang
Abstract
Slender rod insertion arises in precision manufacturing, where millimetre scale diameter and tight clearances demand accurate perception and control. Conventional peg-in-hole methods assume a rigid object whose tip pose is fixed relative to the gripper. This assumption breaks down for a high aspect ratio rod, which can bend during manipulation, making its tip motion dependent on the rod configuration, grasp, material properties, and contact. We present RodForesight, a learning framework that factorises the task into two stages: 1) coarse approaching, which uses visual servoing to map diverse initial configurations into a compact near hole hand-off region; and 2) predictive insertion, which performs fine alignment and completes the insertion. It is worth noting that the two stages can be wrapped into an end-to-end design. During insertion, a diffusion policy generates candidate action chunks, while an action conditioned world model predicts their effects on rod-hole alignment. This pre-execution evaluation enables RodForesight to select the best action chunk based on predicted tilt and radial errors before execution. Experiments investigate the performance of different stages and the end-to-end setting, where RodForesight improves the success rate from 88.9% to 96.7%, compared to baseline methods such as diffusion policy.