UniMorphGrasp: Diffusion Model with Morphology-Awareness for Cross-Embodiment Dexterous Grasp Generation

TL;DR

UniMorphGrasp tackles cross-embodiment dexterous grasping by learning to generate grasps conditioned on both object geometry and hand morphology. It maps heterogeneous hands to a unified canonical pose and uses a Graphormer-based morphology encoder to produce graph-structured features that condition a diffusion-based grasp generator. A morphology-aware loss explicitly weights joints by their position in the kinematic tree and includes physics-based penalties to enforce stability and collision avoidance. The results show state-of-the-art performance on multiple benchmarks, strong zero-shot generalization to unseen morphologies, and practical real-world validation, underscoring the method's potential for scalable cross-embodiment grasp deployment.

Abstract

Cross-embodiment dexterous grasping aims to generate stable and diverse grasps for robotic hands with heterogeneous kinematic structures. Existing methods are often tailored to specific hand designs and fail to generalize to unseen hand morphologies outside the training distribution. To address these limitations, we propose \textbf{UniMorphGrasp}, a diffusion-based framework that incorporates hand morphological information into the grasp generation process for unified cross-embodiment grasp synthesis. The proposed approach maps grasps from diverse robotic hands into a unified human-like canonical hand pose representation, providing a common space for learning. Grasp generation is then conditioned on structured representations of hand kinematics, encoded as graphs derived from hand configurations, together with object geometry. In addition, a loss function is introduced that exploits the hierarchical organization of hand kinematics to guide joint-level supervision. Extensive experiments demonstrate that UniMorphGrasp achieves state-of-the-art performance on existing dexterous grasp benchmarks and exhibits strong zero-shot generalization to previously unseen hand structures, enabling scalable and practical cross-embodiment grasp deployment.

Figures (15)

• Figure 1: We present UniMorphGrasp, a diffusion model with morphology-awareness that can generate diverse cross-embodiment grasps and generalize to novel morphologies.
• Figure 2: (Left) The overview of our proposed UniMorphGrasp 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 canonical hand pose format), 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 morphology-aware denoising model, where the iterative process is conditioned on both the morphology representation and the point cloud representation extracted via a Point Transformer zhao2021pointtransformer. The entire framework is trained based on a morphology-aware loss function. (Right) The structure of our morphology-aware denoising model, which is conditioned on the encoded morphology and the point cloud representations via cross-attention.
• Figure 3: The structure of our morphology encoder. For each joint, we extract its child link's geometric properties, joint limits, origin, and axis to form the joint morphology. The morphologies are embedded into tokens, then processed by a Graphormer ying2021graphormer encoder to obtain morphology representations, where the attention mechanism is biased by the hand's kinematic structure and the active joint mask.
• Figure 4: Visualizations of cross-embodiment grasps synthesized by UniMorphGrasp. Two viewing angles are presented for each grasp.
• Figure 5: Visualizations of ablation study on 1) effectiveness of morphology encoding and 2) zero-shot grasp generalization to novel hand morphologies based on the Shadow Hand. (a) w/o morphology encoding; (b) w/ morphology encoding; (c) w/ morphology encoding and Graphormer; (d) w/ morphology encoding, Graphormer, and morphology-aware loss; (e)-(g) Altered fingers.