Articulated Object Manipulation Using Online Axis Estimation with SAM2-Based Tracking

Xi Wang, Tianxing Chen, Tianling Xu, Yiting Fu, Ziqi He, BingYu Yang, Qiangyu Chen, KailunSu

Keywords: Articulated Object Manipulation, Interactive Perception, Online Axis Estimation, Point-cloud Tracking and Segmentation

TL;DR: We propose a closed-loop pipeline integrates interactive perception with online axis estimation via video segmentation model for articulated object manipulation.

Abstract: We propose a closed-loop pipeline for articulated object manipulation. We first adopt any interactive perception technique to induce slight object movements, and track the evolving manipulation process. We then segment out the point cloud of the articulated object using Segment Anything Model 2, and estimate axes to guide subsequent robotic action. Experiments show that, our method outperforms solely interactive perception methods in tasks requiring precise axis-based control.

Submission Number: 7