MultiGripperGrasp
收藏资源简介:
MultiGripperGrasp是由美国德克萨斯大学达拉斯分校计算机科学系的研究团队创建的大规模机器人抓取数据集。该数据集包含3040万种抓取方式,涵盖11种不同类型的夹具(从两指夹具到五指夹具,包括人手)对345个物体的抓取。抓取数据通过Isaac Sim仿真验证,分为成功与失败抓取,并记录物体脱落时间以衡量抓取质量。数据集的独特之处在于夹具的抓取姿势统一标准化,可实现抓取方式在不同夹具间的迁移,显著提升成功抓取数量。该数据集旨在推动机器人通用抓取规划与跨夹具抓取迁移的研究,助力机器人在复杂环境中实现高效、精准的物体操作。
MultiGripperGrasp is a large-scale robotic grasping dataset created by a research team from the Department of Computer Science at the University of Texas at Dallas, USA. This dataset contains 30.4 million grasping configurations, covering 11 distinct types of grippers (ranging from two-fingered grippers to five-fingered grippers, including human hands) performing grasps on 345 distinct objects. All grasping data is validated via Isaac Sim simulations, categorized into successful and failed grasps, with the object drop time recorded to quantify grasping quality. A unique feature of this dataset is that the grasping poses of the grippers are uniformly standardized, enabling the transfer of grasping strategies across different grippers and significantly increasing the number of successful grasps. This dataset aims to advance research in robotic general grasping planning and cross-gripper grasp transfer, and facilitate robots to perform efficient and precise object manipulation in complex environments.




