Reproduce “Enabling a Unified Cross-Domain Representation for Two-Finger Gripper Manipulation via Interaction-Centric Modeling” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Guanlin Li, Shifeng Bao, Yihan Zhao, Haitao Shen, Haoyang Li, Chen Zhao, Tong Yang, Jie Tang and 1 more

The method represents manipulation scenes in a gripper-centered frame, focusing on the gripper, held object, and next target. A vision-language model, segmentation and tracking, depth, artificial potential fields, point clouds, and a transformer predict actions for two-finger grippers. On 25 RoboTwin2.0 tasks, it reaches 46.1% randomized success versus 7.5% for DP3-W, while Clean favors the baseline, 73.1% versus 66.3%.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.