ankile/square-01a-ours-sobol-r1-diagnostic-rollouts

该数据集是一个机器人数据集，使用LeRobot工具创建，专门针对panda机器人。数据集包含100个episodes，总计22484帧，帧率为20fps，仅用于训练。数据以parquet文件格式存储，视频以mp4格式存储。数据集的结构包括观测状态（如末端执行器位置、四元数姿态、夹爪位置）、动作（如末端执行器位置和旋转变化、夹爪动作）、环境状态（如螺母相对于末端执行器的位置和姿态）、以及其他元数据（如剩余步骤数、来源ID、成功标志、有效性标志、奖励、完成标志、初始模拟关节位置和速度、时间戳、帧索引、episode索引、索引和任务索引）。这些数据可用于机器人学习任务，如强化学习或模仿学习。This dataset is a robotics dataset created using LeRobot, specifically for the panda robot. It contains 100 episodes with a total of 22484 frames at 20 fps, intended for training only. The data is stored in parquet files, and videos are stored in mp4 format. The dataset structure includes observation states (e.g., end-effector position, quaternion orientation, gripper positions), actions (e.g., delta end-effector position and rotation, gripper action), environment states (e.g., nut position and orientation relative to the end-effector), and other metadata (e.g., steps to go, source ID, success flag, validity flag, reward, done flag, initial simulation joint positions and velocities, timestamp, frame index, episode index, index, and task index). This data can be used for robotics learning tasks such as reinforcement learning or imitation learning.