Touch2Insert: Zero-Shot Peg Insertion by Touching Intersections of Peg and Hole
Masaru Yajima, Yuma Shin, Rei Kawakami, Asako Kanezaki, Kei Ota
AI summary
Problem
Reliable insertion of industrial connectors requires sub-millimeter precision under uncertainty, but vision-based methods fail with occlusion and learning-based policies struggle to generalize to unseen geometries. Existing tactile approaches rely on task-specific training or iterative error compensation, making them slow and inefficient for complex shapes.
Approach
The framework reconstructs 3D cross-sectional shapes of the peg and hole from tactile images, filters and projects them to 2D, and estimates their relative SE(2) pose via multi-initialization ICP registration, enabling direct insertion with a stiffness controller.
Key results
- Sub-millimeter pose estimation accuracy across three connector types in simulation
- 86.7% average insertion success rate on a real robot for diverse industrial connectors
- Zero-shot generalization to unseen connector geometries without task-specific training or CAD priors
- Elimination of exploratory spiral searches through direct stiffness-controlled insertion
Why it matters
It provides a robust, vision-independent solution for high-precision robotic assembly, enabling reliable insertion of complex industrial connectors in real-world settings where vision fails.
Abstract
Reliable insertion of industrial connectors remains a central challenge in robotics, requiring sub-millimeter pre- cision under uncertainty and often without full visual access. Vision-based approaches struggle with occlusion and limited generalization, while learning-based policies frequently fail to transfer to unseen geometries. To address these limitations, we leverage tactile sensing, which captures local surface geometry at the point of contact and thus provides reliable informa- tion even under occlusion and across novel connector shapes. Building on this capability, we present Touch2Insert, a tactile- based framework for arbitrary peg insertion. Our method re- constructs cross-sectional geometry from high-resolution tactile images and estimates the relative pose of the hole with respect to the peg in a zero-shot manner. By aligning reconstructed shapes through registration, the framework enables insertion from a single contact without task-specific training. To evaluate its performance, we conducted experiments with three diverse connectors in both simulation and real-robot settings. The results indicate that Touch2Insert achieved sub-millimeter pose estimation accuracy for all connectors in simulation, and attained an average success rate of 86.7% on the real robot, thereby confirming the robustness and generalizability of tactile sensing for real-world robotic connector insertion.