CarlosMunoz0/vla-mujoco-so101-take_out_box-mouse-v2.1

该数据集是一个用于机器人学研究的模拟数据集，使用LeRobot工具创建。它基于Mujoco模拟环境，针对so101机器人模型，设计用于单一任务（可能涉及取物或操作）。数据集包含100个完整的episodes，总计74994帧数据，帧率为30fps。数据以parquet文件格式存储，总数据大小为100MB，视频文件大小为200MB。数据集结构包括动作数据（action，6维浮点数组，表示机器人关节位置和夹爪位置）、状态观测数据（observation.state，13维浮点数组，包括关节位置、末端执行器位置和四元数姿态），以及多个摄像头图像数据：realsense（RGB图像）、realsense_depth（深度图像）、realsense_depth_vis（深度可视化图像）和wrist_cam（腕部摄像头图像），所有图像分辨率为480x640，3通道，视频编码为av1。此外，还包括时间戳、帧索引、episode索引等元数据。数据集仅包含训练分割（train: 0:100），适用于机器人控制、视觉感知和强化学习等任务。

This is a simulation dataset for robotics research, created using the LeRobot toolkit. It is based on the Mujoco simulation environment and targets the so101 robot model, designed for a single task that may involve object picking or manipulation. The dataset contains 100 full episodes, totaling 74,994 frames at a frame rate of 30 fps. The data is stored in Parquet file format, with a total size of 100 MB, while the accompanying video files have a total size of 200 MB. The dataset structure includes action data (action, a 6-dimensional floating-point array representing robot joint positions and gripper positions), state observation data (observation.state, a 13-dimensional floating-point array including joint positions, end-effector positions and quaternion poses), as well as multiple camera image datasets: realsense (RGB image), realsense_depth (depth image), realsense_depth_vis (depth visualization image) and wrist_cam (wrist camera image). All images have a resolution of 480×640 with 3 channels and use AV1 video encoding. Additionally, the dataset includes metadata such as timestamps, frame indices, episode indices and other relevant information. The dataset only provides a training split (train: 0:100), and is suitable for tasks such as robot control, visual perception and reinforcement learning.