EAGG: Embodiment-Aligned Grasp Generation via Geometry-Aware Graph Conditioning

2026-06-16

AI Summary

Existing grasp generation methods struggle to generalize across end-effectors with diverse topologies and contact characteristics—such as parallel-jaw grippers and dexterous hands—due to their reliance on fixed actuator types or static identity encodings. This work proposes EAGG, the first unified generative framework that explicitly aligns end-effector structure by modeling it with a topology-aware graph. Leveraging a frozen cognitive backbone, joint states are mapped into geometry-aware tokens, and an iterative geometric injection mechanism in a low-dimensional control space dynamically synchronizes morphological changes. Evaluated on the MultiGripperGrasp benchmark, EAGG achieves an average grasp success rate of 56.17%—only 1.10 percentage points below specialized models—while enabling zero-shot transfer and fine-tuning. It reduces the median contact distance to 0.189 cm, significantly enhancing cross-end-effector generalization.

Abstract

Cross-end-effector grasp generation seeks a unified model that generalizes across objects and across embodiments ranging from parallel grippers to dexterous end effectors. Existing grasp generators are typically designed for a fixed embodiment or encode embodiment identity with a static descriptor, which weakens transfer when topology, actuation coupling, and contact geometry differ substantially. We present EAGG, an embodiment-aligned grasp generator that represents each embodiment with a topology-aware end-effector graph and an embodiment-specific low-dimensional end-effector control space. A frozen end-effector-cognition backbone converts the current articulated state into geometry-aware tokens that act as a reusable morphology prior, and iterative geometry injection refreshes these tokens throughout sampling so that conditioning remains synchronized with the evolving end-effector geometry. On the MultiGripperGrasp benchmark, EAGG reaches 56.17% average success across six training end effectors, remaining within 1.10 percentage points of specialized training while preserving transfer to finetuning and zero-shot end effectors. Iterative geometry injection further reduces the pooled median contact distance from 0.239 cm to 0.189 cm. These results show that cross-end-effector grasp generation is strengthened by aligning embodiment structure inside a shared generator rather than suppressing embodiment differences. Code is available at https://github.com/wanhaoniu/EAGG.

Problem

• cross-end-effector grasp generation
• embodiment generalization
• grasp synthesis
• end-effector morphology
• contact geometry

Innovation

• embodiment-aligned
• geometry-aware graph
• iterative geometry injection
• cross-end-effector grasp generation
• morphology prior