CarlosMunoz0/vla-mujoco-so101-cube_on_tray-mouse-v2

该数据集使用LeRobot工具创建，专注于机器人技术领域。数据集包含40个训练片段，总计29999帧，帧率为30fps。数据特征包括：6维浮点型动作向量（控制机器人关节位置和夹爪位置）、13维浮点型状态观测向量（包含关节位置、末端执行器位置和四元数姿态）、以及来自RealSense和腕部摄像头的视频观测数据（RGB和深度图像，分辨率480x640）。数据以Parquet格式存储，适用于机器人模仿学习或强化学习任务。

This dataset was created using the LeRobot tool and focuses on the robotics domain. It contains 40 training episodes with a total of 29,999 frames at 30 fps. Features include: a 6-dimensional float32 action vector (controlling robot joint positions and gripper position), a 13-dimensional float32 state observation vector (including joint positions, end-effector position, and quaternion orientation), and video observations from RealSense and wrist cameras (RGB and depth images at 480x640 resolution). Data is stored in Parquet format and is suitable for robot imitation learning or reinforcement learning tasks.