GRAFT: Graph-Based Affordance Transfer via Part Correspondence

2026-06-23

AI Summary

This work addresses the limitation of existing zero-shot robotic manipulation methods that over-rely on semantic cues while neglecting geometric similarity. To overcome this, the authors propose a geometry-aware graph correspondence framework that constructs part-level object graphs and integrates instance- and vertex-level descriptors to align functional parts and propagate contact points across objects. By unifying geometric and functional information within a single graph representation—marking the first such integration in this domain—the method enables high-fidelity manipulation transfer to unseen objects from just a single demonstration. This approach effectively circumvents the bottleneck of semantic retrieval methods that ignore fine-grained geometric details, significantly improving success rates in zero-shot manipulation tasks.

Abstract

Generalizing robotic manipulation to unseen objects remains challenging, as learning-based approaches require many demonstrations and fail in few-shot settings. Prior work transfers affordances through semantic retrieval, but semantics alone neglect geometric similarity, which is critical for manipulation. We propose GRAFT, a geometry-aware correspondence framework for zero-shot manipulation transfer using only one demonstration per object. Objects are represented as part-based graphs, where part-level descriptors support global instance retrieval and part correspondence, and vertex-level descriptors enable fine-grained contact point matching. For an unseen object, our method first retrieves the most functionally and geometrically similar instance from the demonstration buffer with aligned functional parts, and finally propagates the contact points through point-wise correspondence.

Problem

• affordance transfer
• zero-shot manipulation
• geometric similarity
• part correspondence
• unseen objects

Innovation

• zero-shot manipulation
• part-based graph
• geometric correspondence
• affordance transfer
• contact point matching