LightwheelAI/Lightwheel-Tasks-Double-Piper

该数据集包含130个Lightwheel-Libero-Tasks，使用双Piper机器人在LW-BenchHub环境中收集。机器人配置为DoublePiper-Abs，使用DoublePiperAbsEnvCfg继承自DoublePiperEnvCfg。数据集记录了机器人的状态（16维关节位置）和动作（12维，包括左右臂各5自由度及左右夹持器各1自由度）。数据集总共有6500个片段，6010166帧，50fps，包含训练集。数据以parquet格式存储，包含左右手和第一人称视角的视频数据（480x640x3，h264编码）。

This dataset includes 130 Lightwheel-Libero-Tasks, collected using the double-piper robot in environments provided by LW-BenchHub. The robot configuration used is DoublePiper-Abs, using DoublePiperAbsEnvCfg inherits from DoublePiperEnvCfg. It records robot state (16-dim joint positions) and actions (12-dim, including 5-DOF left/right arms and 1-DOF left/right grippers). The dataset contains 6500 episodes, 6010166 frames at 50fps, with a training split. Data is stored in parquet format with left-hand, right-hand and first-person view videos (480x640x3, h264 encoded).