DiffMorphGrasp: A Diffusion Framework Conditioned on Hand Morphology for Cross-Embodiment Grasp Generation (CoRL 2026)

Zhiyuan Wu1, Xiangyu Zhang1, Zhuo Chen1, Jiankang Deng2, Rolandos Alexandros Potamias2, ✉, Shan Luo1, ✉
1King's College London, 2Imperial College London, ✉Corresponding Authors
Preprint Code (Coming soon)

Abstract

As robots are deployed across diverse environments, manipulation tasks require robot hands with different morphologies to satisfy varying task requirements, hardware capabilities, and cost constraints. However, transferring grasping policies learned on one robot hand to another with a different morphology remains a fundamental challenge due to variations in kinematic structures and degrees of freedom across embodiments. To address these limitations, we propose \textbf{DiffMorphGrasp}, the first diffusion-based framework that incorporates hand morphology information into the generative process for cross-embodiment dexterous grasping. The proposed approach maps grasps from diverse robotic hands into a reference hand pose, providing a common space for learning. Grasp generation is then continuously conditioned on structured representations of hand kinematics, encoded as graphs derived from hand configurations, together with object geometry. In addition, we introduce a loss function that computes joint-level errors over the hand kinematic structure to constrain the generated grasps during training. To evaluate DiffMorphGrasp, we construct a new benchmark for assessing grasp generation performance across robot hands with different morphologies. Extensive experiments demonstrate that DiffMorphGrasp achieves strong zero-shot generalization to unseen hand morphologies, enabling scalable cross-embodiment grasp deployment.

Pipeline Overview

DiffMorphGrasp Pipeline
(Left) The overview of our proposed DiffMorphGrasp for cross-embodiment dexterous grasp generation. Given an object point cloud and a target hand morphology extracted from its URDF specification (mapped to a pre-defined reference hand pose), we employ a morphology encoder to extract morphology representations from the hand's joint structure. The hand pose (noised via a diffusion scheduler in training) is embedded through a linear layer, and concatenated with its active joint mask embedding to obtain the hand pose representation. This representation is then processed through a morphological denoising model, where the iterative process is conditioned on both the morphology representation and the point cloud representation extracted via a Point Transformer. The entire framework is trained based on a morphological loss function. (Right) The structure of our morphological denoising model, which is conditioned on the encoded morphology and the point cloud representations via cross-attention.

Method Performance

MultiDex Performance
DiffMorphGrasp can generate stable and diverse grasps for cross-embodiment dexterous hands.
More MultiDex Results
More results generated by our DiffMorphGrasp on the MultiDex dataset.
Quantitative Results 1
Quantitative comparison of our DiffMorphGrasp (w/. and w/o. the morphology-aware loss) with different cross-embodiment dexterous grasp synthesis baselines across three robotic hands from three to five fingers: Barrett, Allegro, and Shadow hand.
Qualitative Comparison
Qualitative comparison with baselines 1) GenDexGrasp and 2) DRO-Grasp, where our results demonstrate superior surface conformity and stable form-closure.

Zero-Shot Generalization to Novel Hand Morphologies

Generalization 1
DiffMorphGrasp can generalize to novel hand morphologies in a zero-shot way, including removing, lengthening, shortening, and replacing fingers.

Zero-Shot Generalization to Unseen Hands

Zero-Shot Generalization to Unseen Hands
DiffMorphGrasp can generalize to completely unseen robotic hands in a zero-shot way.

Cross-Dataset Results

We conduct cross-dataset evaluations on the Multi-GraspLLM and Objaverse datasets to evaluate the zero-shot generalization capability of our model.

Multi GraspLLM
Visualizations of cross-embodiment grasps synthesized by DiffMorphGrasp on the Multi-GraspLLM dataset.
Objaverse Results
Visualizations of cross-embodiment grasps synthesized by DiffMorphGrasp on the Objaverse dataset.
Quantitative Results 2
Cross-dataset zero-shot generalization results. We evaluate models trained on MultiDex directly on unseen datasets: Multi-GraspLLM and Objaverse.

Real-World Experiments

We validate DiffMorphGrasp in real-world scenarios using a UR5e arm equipped with a Leap Hand.

Real-World Grasping Demonstrations

Real-world grasping demonstrations on the Leap Hand.

Quantitative Results 3
Quantitative real-world evaluation on the Leap Hand. We report the success rate over 10 attempts for eight objects from the YCB dataset.
Grasping Strategies from Diverse Directions
Grasping strategies from diverse directions.
Reactive Grasping in Dynamic Environment
Reactive grasping in dynamic environments.