MorphArtGrasp: Morphology-Aware Cross-Embodiment Dexterous Hand Articulation Generation for Grasping

MorphArtGrasp is a new AI framework that enables dexterous robotic hands to grasp objects across different hand designs without extensive retraining. The system achieves 91.9% success rate in simulation and 87% in real-world tests by using morphology-aware learning to adapt grasping strategies to different robotic hand configurations.

• →MorphArtGrasp enables cross-embodiment grasp generation, allowing different robotic hands to learn grasping without hand-specific training datasets.
• →The framework achieves 91.9% grasp success rate in simulation with inference time under 0.4 seconds per grasp.
• →Few-shot adaptation to unseen hands achieves 85.6% simulation success and 87% real-world success rates.
• →The system uses eigengrasp-based approach with morphology embeddings to generalize across different dexterous hand designs.
• →Kinematic-Aware Articulation Loss emphasizes fingertip-relevant motions for improved grasping performance.