GRIP: A General Robotic Incremental Potential Contact Simulation Dataset for Unified Deformable-Rigid Coupled Grasping
收藏资源简介：
GRIP是一个通用的机器人增量潜在接触模拟数据集，用于统一的变形-刚性耦合抓取。该数据集利用优化的增量潜在接触（IPC）模拟器进行多环境数据生成，实现了高达48倍的加速，同时确保了高效、无交叉和无反转的模拟，适用于柔性夹持器和变形物体。我们的全自动管道生成并评估了1200个物体和100,000个抓取姿势的多样化抓取交互，包括软性和刚性夹持器。GRIP数据集支持神经抓取生成和应力场预测等应用。
GRIP is a general-purpose robot incremental potential contact simulation dataset designed for unified deformation-rigid coupled grasping. The dataset employs an optimized incremental potential contact (IPC) simulator to generate multi-environment data with a remarkable 48x acceleration, while ensuring efficient, non-overlapping, and non-inverted simulations. It is suitable for both flexible grippers and deformable objects. Our fully automated pipeline generates and evaluates diverse grasping interactions for 1200 objects and 100,000 grasping poses, including both soft and rigid grippers. The GRIP dataset supports applications such as neural grasping generation and stress field prediction.
GRIP数据集概述
数据集简介
- 名称: GRIP: A General Robotic Incremental Potential Contact Simulation Dataset for Unified Deformable-Rigid Coupled Grasping
- 目的: 提供大规模抓取数据集，支持柔性夹具和软性操作对象的通用抓取模型开发
- 特点:
- 使用优化的IPC模拟器进行多环境数据生成
- 实现最高48倍加速的无交叉、无反转模拟
- 包含1,200个对象和100,000个抓取姿态
- 支持软性和刚性夹具
数据集发布计划
- 已发布:
finray数据集
- 待发布:
bifinray数据集leaphand数据集bileaphand数据集- 统一数据生成管道
- 数据集可视化工具包
数据集下载
-
下载地址: https://ucla.box.com/s/zc1fxvv2cj5ynodtoirykojggqw051g7
-
文件结构:
grasp-dataset/ └── release_version/ ├── finray_soft_837 ├── finray_soft_837.part.aa ├── finray_soft_837.part.ab ├── ...
-
解压说明: 需合并分块文件后解压，需>1T磁盘空间
数据集结构
-
子集:
finray_soft_837/finray_rigid_710
-
文件命名格式:
<hand_type><object_name><material_type>
-
H5文件内容:
object_config: 对象物理属性- 对象类型、密度、杨氏模量、泊松比、摩擦系数等
trajectories: 模拟数据- 20个成功抓取轨迹，每轨迹100帧
- 包含夹具状态、位姿、网格顶点、接触点等数据
引用
bibtex @misc{ma2025gripgeneralroboticincremental, title={GRIP: A General Robotic Incremental Potential Contact Simulation Dataset for Unified Deformable-Rigid Coupled Grasping}, author={Siyu Ma and Wenxin Du and Chang Yu and Ying Jiang and Zeshun Zong and Tianyi Xie and Yunuo Chen and Yin Yang and Xuchen Han and Chenfanfu Jiang}, year={2025}, eprint={2503.05020}, archivePrefix={arXiv}, primaryClass={cs.RO}, url={https://arxiv.org/abs/2503.05020}, }
许可证
- 类型: CC BY-NC 4.0
- 链接: https://creativecommons.org/licenses/by-nc/4.0/
