UniFucGrasp: Human-Hand-Inspired Unified Functional Grasp Annotation Strategy and Dataset for Diverse Dexterous Hands

TL;DR

The paper tackles the lack of large-scale, multi-hand functional grasp data by proposing UniFucGrasp, a unified human-to-robot hand mapping framework that supports both fully-actuated and under-actuated dexterous hands via a sparse W mapping and RTJ control, coupled with geometry-based force-closure validation.It introduces a large-scale UniFucGrasp dataset (over 100k functional grasp annotations across 1,108 objects in 21 categories) and a functional gesture generation model conditioned on hand–object point clouds, leveraging a CVAE and Transformer for stable, human-like grasps.Experiments in MuJoCo and real-world setups demonstrate improved functional manipulation accuracy and grasp stability, with better generalization across multiple robotic hands and tasks such as bottle handling and pouring.These contributions reduce annotation cost, enable cross-hand portability, and advance practical dexterous manipulation research by aligning grasp gestures with functional intents using human priors.

Abstract

Dexterous grasp datasets are vital for embodied intelligence, but mostly emphasize grasp stability, ignoring functional grasps needed for tasks like opening bottle caps or holding cup handles. Most rely on bulky, costly, and hard-to-control high-DOF Shadow Hands. Inspired by the human hand's underactuated mechanism, we establish UniFucGrasp, a universal functional grasp annotation strategy and dataset for multiple dexterous hand types. Based on biomimicry, it maps natural human motions to diverse hand structures and uses geometry-based force closure to ensure functional, stable, human-like grasps. This method supports low-cost, efficient collection of diverse, high-quality functional grasps. Finally, we establish the first multi-hand functional grasp dataset and provide a synthesis model to validate its effectiveness. Experiments on the UFG dataset, IsaacSim, and complex robotic tasks show that our method improves functional manipulation accuracy and grasp stability, demonstrates improved adaptability across multiple robotic hands, helping to alleviate annotation cost and generalization challenges in dexterous grasping. The project page is at https://haochen611.github.io/UFG.

Figures (11)

• Figure 1: Mapping natural human motions to an anthropomorphic hand model and unified control across representative robotic hands (ShadowHand, InspireHand), and HnuHand 10011809. Visualization of functional grasps from UniFucGrasp dataset, including interactions with bottle, camera, mug, and drill.
• Figure 2: Illustration of our annotation strategy for transferring human motions to robotic hands, with the left showing capture, the top depicting gesture mapping, the bottom showing pose collection, and the right displaying the resulting functional grasps.
• Figure 3: Flowchart of data acquisition for mapping joint angles from the human index finger to the robotic index finger.
• Figure 4: For each contact $i$$(\mathbf{n}_i, \mathbf{p}_i)$, we show its six friction cone directions $\mathbf{w}_{i,j}$ and the reduced wrench spaces (blue regions).
• Figure 5: Functional gesture generation for diverse robotic hands, with point cloud feature extraction, CVAE Sohn_Yan_Lee_2015 latent encoding and sampling, and MLP decoder for gesture parameters.