bottle-opening-combined-delta-stats

该数据集是一个机器人操作数据集，专门针对开瓶任务设计。数据集包含使用Anvil OpenArm双机械臂平台收集的多模态数据：1）视觉数据：从四个摄像头视角（腰部、胸部、左腕、右腕）采集的RGB视频，分辨率480x640，帧率30fps；2）机器人状态数据：16维关节状态（左右臂各7个关节加1个手指关节）、关节动作指令、关节速度和关节力矩；3）元数据：时间戳、帧索引、片段索引等。数据集共包含170个训练片段，总计133,761帧数据，专门用于机器人学习开瓶动作的模仿学习或强化学习研究。This dataset is a robotic manipulation dataset specifically designed for bottle opening tasks. It contains multimodal data collected using the Anvil OpenArm dual robotic arm platform: 1) Visual data: RGB videos captured from four camera perspectives (waist, chest, left wrist, right wrist) with a resolution of 480x640 and a frame rate of 30fps; 2) Robot state data: 16-dimensional joint states (7 joints per arm plus 1 finger joint), joint action commands, joint velocities, and joint torques; 3) Metadata: timestamps, frame indices, segment indices, etc. The dataset includes 170 training segments, totaling 133,761 frames, and is specifically intended for imitation learning or reinforcement learning research in robotic bottle opening actions.