DORLR/teleop_isaaclab_ur5_udsci_success_36_demo
收藏资源简介:
该数据集使用LeRobot创建,专注于机器人学领域,特别是UR5机器人操作任务。数据集涉及环境UR5-Deformable-StackedCuboid-Insertion-v0,旨在模拟可变形堆叠长方体插入操作。数据集包含2个episodes,总计4965帧,覆盖1个任务。数据结构以parquet文件格式组织,包括视频文件(MP4格式)和特征数据。特征包括时间戳、帧索引、episode索引、任务索引,以及观察数据:来自cam_side2和cam_color摄像头的图像(分辨率512x512,3通道)、机器人关节状态(7个关节:肩部平移、肩部提升、肘部、腕部1、腕部2、腕部3、手指关节)、笛卡尔状态(末端执行器位置和四元数:eef_x, eef_y, eef_z, eef_qx, eef_qy, eef_qz, eef_qw)和动作(7个关节控制)。数据集用于训练和评估机器人控制算法,支持强化学习或模仿学习应用。
This dataset was created using LeRobot. It focuses on robotics, specifically UR5 robot manipulation tasks. The dataset involves the environment UR5-Deformable-StackedCuboid-Insertion-v0, simulating deformable stacked cuboid insertion operations. It contains 2 episodes, totaling 4965 frames, covering 1 task. The data structure is organized in parquet file format, including video files (MP4 format) and feature data. Features include timestamp, frame index, episode index, task index, and observations: images from cam_side2 and cam_color cameras (resolution 512x512, 3 channels), robot joint state (7 joints: shoulder_pan_joint, shoulder_lift_joint, elbow_joint, wrist_1_joint, wrist_2_joint, wrist_3_joint, finger_joint), Cartesian state (end-effector position and quaternion: eef_x, eef_y, eef_z, eef_qx, eef_qy, eef_qz, eef_qw), and actions (7 joint controls). The dataset is used for training and evaluating robot control algorithms, supporting reinforcement learning or imitation learning applications.




