ankile/square-d1-01a-ours-sobol-r1-freecf
收藏资源简介:
这是一个使用LeRobot创建的机器人学习数据集,专门用于机器人操作任务。数据集包含100个训练片段(episodes),总计20867帧,数据以parquet格式存储。机器人类型为panda,帧率为20fps。数据集特征丰富,包括观察状态(如末端执行器位置、四元数姿态和夹爪位置)、环境状态(如螺母与末端执行器的相对位置和姿态、螺母和钉的绝对位置)、动作(如末端执行器的位置和旋转增量以及夹爪动作),以及辅助信息如来源ID、成功标志、干预标志、有效标志、奖励、完成标志、时间戳、帧索引、片段索引、索引和任务索引。数据集主要用于训练和评估机器人控制算法,特别是涉及精细操作(如装配任务)的场景。
This is a robotic learning dataset created using LeRobot, specifically designed for robotic manipulation tasks. It contains 100 training episodes, totaling 20867 frames, and is stored in parquet format. The robotic platform is Panda, with a frame rate of 20 fps. The dataset has rich features including observation states (such as end-effector position, quaternion orientation and gripper position), environmental states (such as the relative position and orientation between the nut and the end-effector, as well as the absolute positions of the nut and peg), actions (such as the position and rotation increments of the end-effector and gripper movements), along with auxiliary information such as source ID, success flag, intervention flag, valid flag, reward, completion flag, timestamp, frame index, episode index, index and task index. This dataset is primarily used for training and evaluating robotic control algorithms, especially in scenarios involving fine manipulation tasks such as assembly operations.




