Robots with multi-fingered hands often struggle to grasp objects when the hand changes, because grasp models tend to be tuned to a specific gripper. Existing interaction-centric methods also underrepresent local surface geometry and treat contact in an isotropic way. MANGO-Grasp, a new framework from the paper's authors, addresses these issues by representing objects as geometry-oriented 3D Gaussian primitives, which better preserve local surface geometry, and by encoding hands into pretrained descriptors that capture both morphology and kinematics.
Hand-object interaction is modeled with anisotropic Mahalanobis fields, which reflect the directional nature of contact and guide a shared optimization for grasp realization across embodiments. This means the same model can synthesize grasps for different hands without embodiment-specific tuning.
On the CMAP and MultiGripperGrasp benchmarks, MANGO-Grasp outperforms the strongest seen-hand baseline by up to 8.24 percentage points in simulation. It also transfers zero-shot to the unseen SharpaWave hand, improving over the strongest zero-shot baseline by up to 16.57 percentage points, and achieves 86% success in real-world experiments. The code and additional materials will be made available upon publication.