VarunGiridhar3/libero40_libero_object_play

该数据集是通过LeRobot平台创建的机器人操作数据集，专门用于机器人学习任务。数据集包含420个完整的操作序列（episodes），总计117,600帧，覆盖10个不同的任务。数据采集使用panda机器人，记录内容包括：7维浮点动作向量；状态观测，如末端执行器位置（x, y, z）、姿态（aa_x, aa_y, aa_z）和夹爪状态（grip_l, grip_r）；两个图像观测，均为256x256像素的RGB图像，以视频形式存储；以及时间戳、帧索引、episode索引等元数据。数据以parquet文件组织，视频采用av1编码的mp4格式，帧率为10fps。数据集适用于机器人控制、强化学习或模仿学习的研究和应用。This dataset is a robotics manipulation dataset created using the LeRobot platform, designed for robot learning tasks. It contains 420 complete episodes, totaling 117,600 frames, and covers 10 distinct tasks. Data was collected using a panda robot, recording: 7-dimensional float action vectors; state observations including end-effector position (x, y, z), orientation (aa_x, aa_y, aa_z), and gripper states (grip_l, grip_r); two image observations, both 256x256 pixel RGB images stored as videos; and metadata such as timestamps, frame indices, and episode indices. The data is organized in parquet files, with videos in av1-encoded mp4 format at 10fps. The dataset is suitable for research and applications in robot control, reinforcement learning, or imitation learning.