BridgeData V2
收藏资源简介:
BridgeData V2是由加州大学伯克利分校和斯坦福大学等机构合作创建的大型机器人操作行为数据集,包含60,096个轨迹,覆盖24个环境。数据集设计用于支持大规模机器人学习研究,涵盖广泛的技能和环境变量,支持通过目标图像或自然语言指令进行任务调节。创建过程中,数据集通过人工操作和自主收集相结合的方式,确保了数据的多源性和多样性。该数据集适用于多种机器人学习方法,旨在解决机器人技能在不同环境和任务中的泛化问题,推动机器人学习领域的研究进展。
BridgeData V2 is a large-scale robotic manipulation behavior dataset co-developed by institutions including the University of California, Berkeley and Stanford University. It contains 60,096 trajectories covering 24 distinct environments. Designed to support large-scale robotic learning research, this dataset covers a broad spectrum of skills and environmental variables, and enables task conditioning via target images or natural language instructions. During its creation, the dataset was collected through a hybrid approach of manual operation and autonomous data gathering, ensuring the multi-source nature and diversity of the dataset. This dataset is compatible with a variety of robotic learning methods, aiming to address the generalization problem of robotic skills across different environments and tasks, and promote the advancement of research in the field of robotic learning.




