GraspGen-X: Cross-Embodiment 6-DOF Diffusion-based Grasping

2026-05-31

AI Summary

This work addresses the challenge of generalizing six-degree-of-freedom robotic grasping across diverse gripper morphologies and actuation mechanisms, aiming to simultaneously handle novel objects, unseen scenes, and previously unencountered grippers. To this end, the authors propose a diffusion-based grasp generation method that innovatively encodes gripper geometry using sweep volumes and is trained on a large-scale procedurally generated dataset comprising two billion samples. The resulting model constitutes the first six-DoF grasping system capable of zero-shot cross-embodiment generalization. It significantly outperforms existing baselines in simulation and successfully transfers to real-world robotic platforms equipped with novel grippers, thereby demonstrating its effectiveness as a universal grasp prior.

Abstract

We study cross-embodiment 6-DOF robot grasping. Unlike prior works, we require the model not only to generalize to novel objects / scenes but also to novel gripper morphologies and physical grasping processes. Our method extends diffusion model based generative 6-DOF grasping models to condition on the additional gripper's representation. We propose a swept-volume heuristic for encoding the gripper. We train our cross-embodiment model with procedural grippers and a large-scale dataset of 2 Billion grasps. In simulation experiments, our model has the best zero-shot generalization to novel real-world grippers and objects over baseline methods. Our model also serves as a good initialization for fine-tuning to adapt to novel grippers. In ablations, we demonstrate the efficiency of our sweep-volume gripper representation and our procedural gripper training dataset. Last, we show zero-shot generalization to real-world novel grippers for 6-DOF grasping, surpassing baselines in cross-embodiment generalization.

Problem

• cross-embodiment
• 6-DOF grasping
• gripper generalization
• zero-shot grasping
• robotic manipulation

Innovation

• cross-embodiment
• 6-DOF grasping
• diffusion model
• swept-volume representation
• procedural grippers