GeoMatch++: Morphology Conditioned Geometry Matching for Multi-Embodiment Grasping

TL;DR

This paper introduces GeoMatch++, a unified approach that leverages robot morphology through attention mechanisms to improve generalization of dexterous grasping to unseen end-effectors, achieving significant success rate improvements.

Contribution

It proposes a novel morphology-conditioned geometry matching method that enhances grasping generalization across different robot end-effectors.

Findings

Average 9.64% increase in grasp success rate on unseen end-effectors

Effective use of morphology information through attention mechanisms

Improved generalization in multi-embodiment grasping scenarios

Abstract

Despite recent progress on multi-finger dexterous grasping, current methods focus on single grippers and unseen objects, and even the ones that explore cross-embodiment, often fail to generalize well to unseen end-effectors. This work addresses the problem of dexterous grasping generalization to unseen end-effectors via a unified policy that learns correlation between gripper morphology and object geometry. Robot morphology contains rich information representing how joints and links connect and move with respect to each other and thus, we leverage it through attention to learn better end-effector geometry features. Our experiments show an average of 9.64% increase in grasp success rate across 3 out-of-domain end-effectors compared to previous methods.