D_body

该数据集包含了Dreamer代理在视觉设置下为Walker和Cheetah实体生成的轨迹的体现数据集。该数据集是通过在1百万环境步骤中训练Dreamer代理生成的，其中真实状态包括了每个关节的位置和速度信息。数据集规模包含1000条轨迹，涵盖了不同的任务，对于Walker实体包括站立、行走和跑步；对于Cheetah实体则包括跑步、倒跑、翻滚以及反向翻滚。This dataset is an embodied trajectory dataset generated by the Dreamer agent for Walker and Cheetah entities in the visual setup. It was produced by training the Dreamer agent across 1 million environment steps, where the ground-truth states include the position and velocity information of each joint. The dataset consists of 1000 trajectories covering a range of tasks: for the Walker entity, the tasks are standing, walking, and running; for the Cheetah entity, the tasks include running, backward running, tumbling, and backward tumbling.