edmos7/so101_squirtle-trajectory

该数据集是一个机器人学数据集，使用LeRobot创建，专门用于机器人控制任务。数据集包含64个episodes，总计32505帧，覆盖2个不同的任务。数据以parquet文件格式存储，视频文件为mp4格式，帧率为30fps。数据集的特征包括：动作数据（由6个浮点数组成，对应机械臂的6个关节：主肩部平移、主肩部提升、主肘部弯曲、主腕部弯曲、主腕部滚动和主夹爪）、观测状态（同样为6个浮点数，表示机械臂的关节状态）、时间戳、帧索引、episode索引、索引、任务索引，以及来自MAIN_CAM和SEC_CAM的视频观测（分辨率480x640，3通道，编码为av1，无音频）和轨迹图像观测（分辨率480x640，3通道）。数据集适用于机器人学习和控制算法的训练与评估。This dataset is a robotics dataset created using LeRobot, specifically designed for robot control tasks. It contains 64 episodes with a total of 32,505 frames, covering 2 distinct tasks. The data is stored in parquet files, with video files in mp4 format at 30fps. The dataset features include: action data (composed of 6 float values corresponding to the robot arms 6 joints: main shoulder pan, main shoulder lift, main elbow flex, main wrist flex, main wrist roll, and main gripper), observation state (also 6 float values representing the joint states of the robot arm), timestamp, frame index, episode index, index, task index, and video observations from MAIN_CAM and SEC_CAM (resolution 480x640, 3 channels, encoded as av1, without audio) and trajectory image observations (resolution 480x640, 3 channels). The dataset is suitable for training and evaluating robot learning and control algorithms.