# OV-PEAR preprint reports robotic assembly using models of unknown industrial objects

_Published Monday, October 19, 2026 at 10:12 PM EDT · Robotics, Science · Latest · Tier 2 — Notable_

Researchers report a method for reconstructing unknown industrial objects and estimating their position, orientation and scale, with real-robot connector tests achieving 99.67% grasp success and 84.00% assembly success. The OV-PEAR preprint addresses manipulation tasks for which existing methods rely on CAD models, object-specific training data or stable workcells.

The system generates candidate 3D models from color and depth images, aligns them with the robot's coordinate frame, and checks them against observations from multiple viewpoints to suppress invented geometry. It produces reusable object models, but assembly success remains below grasp success in the reported tests.

## Sources

- [arXiv Query: search_query=cat:cs.RO&id_list=&start=0&max_results=30](https://arxiv.org/abs/2610.25840v1)

---
Canonical: https://techandbusiness.org/newswire/4Gjp4j3SwMck9QYHGh913N
Published: 2026-10-20T02:12:16.563Z
Story chronology: 2026-10-19T05:51:41.000Z
Retrieved: 2026-10-20T06:11:28.304Z
Publisher: Tech & Business (techandbusiness.org)