jellyho/aloha_handover_box_joint_pos_rl_orig
收藏资源简介:
该数据集是一个用于机器人学习的数据集,由LeRobot创建。它包含50个训练episodes,总计11829帧和150个视频。数据集以parquet格式存储,特征包括机器人状态观测(14维,对应左右机械臂的6个关节和夹爪位置)、动作(同样14维)、来自三个摄像头的图像观测(背部摄像头分辨率为480x640,左右手腕摄像头分辨率为240x320)、奖励信号、成功标志、时间戳、帧索引、episode索引等。数据以25fps采集,适用于机器人控制、强化学习或模仿学习任务。
This dataset is a robotics learning dataset created by LeRobot. It contains 50 training episodes, totaling 11829 frames and 150 videos. The data is stored in parquet format and includes features such as robot state observations (14 dimensions corresponding to 6 joints and gripper positions for both left and right arms), actions (also 14 dimensions), image observations from three cameras (back camera at 480x640 resolution, left and right wrist cameras at 240x320 resolution), reward signals, success flags, timestamps, frame indices, episode indices, and more. The data was captured at 25fps and is suitable for robotics control, reinforcement learning, or imitation learning tasks.




