ZhaoRunyi/Piper_insert_test_tube_0426

Piper LeRobot数据集是一个机器人操作数据集，专注于双机械臂（左臂和右臂）执行插入试管等任务。数据集由脚本convert_piper_hdf5_to_lerobot.py于2026年4月26日生成，使用aloha机器人类型。数据采集频率为10 FPS，采用全采样模式，以高分辨率摄像头（slave_cam_high）作为主导组进行时间戳对齐。数据集包含状态和动作数据，状态数据涵盖关节位置（各臂6个关节）、夹爪状态（原始格式）、末端执行器位置（3维坐标）和旋转（使用rot6d格式表示），动作数据与状态数据索引顺序一致。图像数据包括来自高分辨率摄像头、左腕摄像头和右腕摄像头的观测图像。数据预处理涉及时间戳对齐、旋转格式转换（从Piper xyz Euler到rot6d）等。数据集旨在支持机器人学习任务，如模仿学习或强化学习，应用于实验室自动化等场景。

The Piper LeRobot dataset is a robotic manipulation dataset focusing on dual-arm (left and right arms) tasks such as inserting test tubes. It was generated by the script convert_piper_hdf5_to_lerobot.py on April 26, 2026, using the aloha robot type. Data is collected at 10 FPS with a full sampling mode, and timestamp alignment is performed with the high-resolution camera (slave_cam_high) as the leading group. The dataset includes state and action data: state data covers joint positions (6 joints per arm), gripper status (raw format), end-effector position (3D coordinates), and rotation (represented in rot6d format), while action data follows the same index order as state. Image data comprises observation images from a high-resolution camera, left wrist camera, and right wrist camera. Preprocessing involves timestamp alignment, rotation format conversion (from Piper xyz Euler to rot6d), etc. The dataset is designed to support robotic learning tasks, such as imitation learning or reinforcement learning, for applications like laboratory automation.