real01c-marker-insert-d1-bc-r0

该数据集是一个由LeRobot工具创建的机器人控制数据集，专门用于机器人学习和强化学习任务。它包含200个训练episodes，总计38680帧，采样频率为15 fps，机器人类型为Franka。数据以parquet文件和视频文件形式存储，总数据大小约100 MB，视频大小约200 MB。数据集特征丰富，包括观测状态（如笛卡尔位置、关节位置、速度、夹爪位置）、动作（如笛卡尔速度、关节速度、夹爪控制）、奖励、完成标志、成功标志、有效性标志等。此外，提供来自两个摄像头（ID: 18650758_left和25916956_left）的图像观测，分辨率为480x640，视频编码为h264。数据集结构支持分块处理，每个chunk包含1000个数据点，适用于机器人控制算法的训练和评估。This dataset is a robotic control dataset created by the LeRobot toolkit, specifically designed for robot learning and reinforcement learning tasks. It contains 200 training episodes, totaling 38,680 frames with a sampling frequency of 15 fps, and the robot type is Franka. The data is stored in parquet files and video files, with a total data size of approximately 100 MB, and the total size of the video files is around 200 MB. The dataset has rich features including observation states (such as Cartesian position, joint position, velocity, and gripper position), actions (including Cartesian velocity, joint velocity, and gripper control), rewards, completion flags, success flags, validity flags, and others. Additionally, image observations from two cameras (ID: 18650758_left and 25916956_left) are provided, with a resolution of 480×640 and the video encoded in h264 format. The dataset structure supports chunked processing, where each chunk contains 1,000 data points, which is suitable for the training and evaluation of robotic control algorithms.

原始信息汇总

数据集概述

• 名称: ankile/real01c-marker-insert-d1-bc-r0
• 许可证: Apache-2.0
• 任务类别: 机器人学 (Robotics)
• 创建工具: 使用 LeRobot 创建
• 机器人类型: Franka

数据集规模

• 总片段数 (episodes): 200
• 总帧数 (frames): 38,680
• 总任务数 (tasks): 1
• 帧率 (FPS): 15
• 数据文件大小: 100 MB
• 视频文件大小: 200 MB
• 数据分割: 训练集 (train) 包含全部 200 个片段 (0:200)

数据集结构

数据以 LeRobot 格式存储，包含 Parquet 数据文件和 MP4 视频文件。

• 数据路径: data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet
• 视频路径: videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4
• 元信息文件: meta/info.json (代码库版本: v3.0, 块大小: 1000)

特征（Features）

该数据集包含多种特征，主要分为以下几类：

观测状态 (observation.state)

特征名	数据类型	形状	字段名称
observation.state	float32	(7,)	cart_pos_x, cart_pos_y, cart_pos_z, cart_rot_x, cart_rot_y, cart_rot_z, gripper_position
observation.state.cartesian_position	float32	(6,)	x, y, z, roll, pitch, yaw
observation.state.joint_position	float32	(7,)	joint_0 至 joint_6
observation.state.joint_velocity	float32	(7,)	joint_0 至 joint_6
observation.state.cartesian_velocity	float32	(6,)	x, y, z, roll, pitch, yaw
observation.state.gripper_position	float32	(1,)	gripper

观测图像 (observation.images)

提供两个相机视角的视频数据。

特征名	分辨率	视频编码	帧率
observation.images.18650758_left	480x640x3 (HWC)	h264	15 FPS
observation.images.25916956_left	480x640x3 (HWC)	h264	15 FPS

动作 (action)

包含多种动作表示形式。

特征名	数据类型	形状	字段名称
action	float32	(7,)	vel_x, vel_y, vel_z, vel_roll, vel_pitch, vel_yaw, gripper_action
action.cartesian_velocity	float32	(6,)	x, y, z, roll, pitch, yaw
action.cartesian_position	float32	(6,)	x, y, z, roll, pitch, yaw
action.joint_velocity	float32	(7,)	joint_0 至 joint_6
action.joint_position	float32	(7,)	joint_0 至 joint_6
action.gripper_position	float32	(1,)	gripper
action.gripper_velocity	float32	(1,)	gripper

其他标量特征

特征名	数据类型	形状	字段名称
steps_to_go	int64	(1,)	steps_to_go
source	int64	(1,)	source_id
success	int64	(1,)	success_flag
is_valid	int64	(1,)	is_valid_flag
reward	float32	(1,)	reward
done	int64	(1,)	done_flag
manifest_idx	int64	(1,)	manifest_idx
pen_x	float32	(1,)	pen_x
pen_y	float32	(1,)	pen_y
pen_yaw	float32	(1,)	pen_yaw
timestamp	float32	(1,)	无
frame_index	int64	(1,)	无
episode_index	int64	(1,)	无
index	int64	(1,)	无
task_index	int64	(1,)	无

引用信息

目前暂无可用的 BibTeX 引用信息。