robotwin-multi
收藏资源简介:
RoboTwin Multi 是一个多具身机器人孪生数据集。该数据集基于19个不同的任务生成,每个任务都包含了8种不同的机器人具身配置。针对每一个“任务-具身配置”组合,数据集中提供了50个演示样本。因此,数据集的总体规模由任务数量、配置数量以及每个组合的演示数量共同决定。需要注意的是,数据集的生成过程有意排除了`put_object_cabinet`这个任务,因为该任务在最终的数据生成运行中未能稳定地产生全部8种具身配置。该数据集适用于机器人学习、模仿学习以及需要研究不同物理形态(具身配置)对任务执行影响的领域。
RoboTwin Multi is a multi-embodied robot twin dataset. The dataset is generated based on 19 different tasks, each containing 8 different robot embodiment configurations. For each task-embodiment configuration combination, the dataset provides 50 demonstration samples. Therefore, the overall scale of the dataset is determined by the number of tasks, configurations, and demonstrations per combination. It should be noted that the dataset generation process intentionally excluded the `put_object_cabinet` task because it failed to stably produce all 8 embodiment configurations in the final data generation run. This dataset is suitable for robot learning, imitation learning, and fields that require studying the impact of different physical forms (embodiment configurations) on task execution.
数据集名称
RoboTwin Multi
许可证
Apache-2.0
数据集描述
多具身机器人数据集,基于19个任务生成,每个任务包含8种具身配置,每个任务-具身配置组合包含50个演示样本。
重要说明
该上传版本特意排除了 put_object_cabinet 任务,因为在最终生成过程中该任务未能稳定地达到8种具身配置。




