ecappiell/multiple_box_experiment_20260527_183717

--- license: apache-2.0 task_categories: - robotics tags: - LeRobot configs: - config_name: default data_files: data/*/*.parquet --- This dataset was created using [LeRobot](https://github.com/huggingface/lerobot). <a class="flex" href="https://huggingface.co/spaces/lerobot/visualize_dataset?path=ecappiell/multiple_box_experiment_20260527_183717"> <img class="block dark:hidden" src="https://huggingface.co/datasets/huggingface/badges/resolve/main/visualize-this-dataset-xl.svg"/> <img class="hidden dark:block" src="https://huggingface.co/datasets/huggingface/badges/resolve/main/visualize-this-dataset-xl-dark.svg"/> </a> ## Dataset Description - **Homepage:** [More Information Needed] - **Paper:** [More Information Needed] - **License:** apache-2.0 ## Dataset Structure [meta/info.json](meta/info.json): ```json { "codebase_version": "v3.0", "fps": 30, "features": { "action": { "dtype": "float32", "names": [ "shoulder_pan.pos", "shoulder_lift.pos", "elbow_flex.pos", "wrist_flex.pos", "wrist_roll.pos", "gripper.pos" ], "shape": [ 6 ] }, "observation.state": { "dtype": "float32", "names": [ "shoulder_pan.pos", "shoulder_pan.current_raw", "shoulder_pan.load_raw", "shoulder_pan.vel_hw", "shoulder_lift.pos", "shoulder_lift.current_raw", "shoulder_lift.load_raw", "shoulder_lift.vel_hw", "elbow_flex.pos", "elbow_flex.current_raw", "elbow_flex.load_raw", "elbow_flex.vel_hw", "wrist_flex.pos", "wrist_flex.current_raw", "wrist_flex.load_raw", "wrist_flex.vel_hw", "wrist_roll.pos", "wrist_roll.current_raw", "wrist_roll.load_raw", "wrist_roll.vel_hw", "gripper.pos", "gripper.current_raw", "gripper.load_raw", "gripper.vel_hw", "shoulder_pan.vel", "shoulder_lift.vel", "elbow_flex.vel", "wrist_flex.vel", "wrist_roll.vel", "gripper.vel", "shoulder_pan.torque_raw", "shoulder_lift.torque_raw", "elbow_flex.torque_raw", "wrist_flex.torque_raw", "wrist_roll.torque_raw", "gripper.torque_raw", "shoulder_pan.current_raw_signed", "shoulder_pan.current_ma_signed", "shoulder_pan.torque_raw_signed", "shoulder_lift.current_raw_signed", "shoulder_lift.current_ma_signed", "shoulder_lift.torque_raw_signed", "elbow_flex.current_raw_signed", "elbow_flex.current_ma_signed", "elbow_flex.torque_raw_signed", "wrist_flex.current_raw_signed", "wrist_flex.current_ma_signed", "wrist_flex.torque_raw_signed", "wrist_roll.current_raw_signed", "wrist_roll.current_ma_signed", "wrist_roll.torque_raw_signed", "gripper.current_raw_signed", "gripper.current_ma_signed", "gripper.torque_raw_signed", "shoulder_pan.filtered/vel_rad", "shoulder_pan.estimated/tau_ext", "shoulder_pan.estimated/tau_var", "shoulder_pan.model/tau_motor", "shoulder_pan.model/tau_gravity", "shoulder_pan.model/tau_friction", "shoulder_pan.model/tau_coriolis", "shoulder_pan.model/tau_inertial", "shoulder_pan.model/tau_model", "shoulder_lift.filtered/vel_rad", "shoulder_lift.estimated/tau_ext", "shoulder_lift.estimated/tau_var", "shoulder_lift.model/tau_motor", "shoulder_lift.model/tau_gravity", "shoulder_lift.model/tau_friction", "shoulder_lift.model/tau_coriolis", "shoulder_lift.model/tau_inertial", "shoulder_lift.model/tau_model", "elbow_flex.filtered/vel_rad", "elbow_flex.estimated/tau_ext", "elbow_flex.estimated/tau_var", "elbow_flex.model/tau_motor", "elbow_flex.model/tau_gravity", "elbow_flex.model/tau_friction", "elbow_flex.model/tau_coriolis", "elbow_flex.model/tau_inertial", "elbow_flex.model/tau_model", "wrist_flex.filtered/vel_rad", "wrist_flex.estimated/tau_ext", "wrist_flex.estimated/tau_var", "wrist_flex.model/tau_motor", "wrist_flex.model/tau_gravity", "wrist_flex.model/tau_friction", "wrist_flex.model/tau_coriolis", "wrist_flex.model/tau_inertial", "wrist_flex.model/tau_model", "wrist_roll.filtered/vel_rad", "wrist_roll.estimated/tau_ext", "wrist_roll.estimated/tau_var", "wrist_roll.model/tau_motor", "wrist_roll.model/tau_gravity", "wrist_roll.model/tau_friction", "wrist_roll.model/tau_coriolis", "wrist_roll.model/tau_inertial", "wrist_roll.model/tau_model", "gripper.filtered/vel_rad", "gripper.estimated/tau_ext", "gripper.estimated/tau_var", "gripper.model/tau_motor", "gripper.model/tau_gravity", "gripper.model/tau_friction", "gripper.model/tau_coriolis", "gripper.model/tau_inertial", "gripper.model/tau_model", "tcp/wrench/fx", "tcp/wrench/fy", "tcp/wrench/fz", "tcp/wrench/tx", "tcp/wrench/ty", "tcp/wrench/tz", "hpi/gripper/tau_ext", "hpi/gripper/q", "hpi/gripper/dq" ], "shape": [ 117 ] }, "observation.images.top": { "dtype": "video", "shape": [ 480, 640, 3 ], "names": [ "height", "width", "channels" ], "info": { "video.height": 480, "video.width": 640, "video.codec": "av1", "video.pix_fmt": "yuv420p", "video.is_depth_map": false, "video.fps": 30, "video.channels": 3, "has_audio": false, "video.g": 2, "video.crf": 30, "video.preset": 12, "video.fast_decode": 0, "video.video_backend": "pyav", "video.extra_options": {} } }, "observation.images.corner": { "dtype": "video", "shape": [ 480, 640, 3 ], "names": [ "height", "width", "channels" ], "info": { "video.height": 480, "video.width": 640, "video.codec": "av1", "video.pix_fmt": "yuv420p", "video.is_depth_map": false, "video.fps": 30, "video.channels": 3, "has_audio": false, "video.g": 2, "video.crf": 30, "video.preset": 12, "video.fast_decode": 0, "video.video_backend": "pyav", "video.extra_options": {} } }, "timestamp": { "dtype": "float32", "shape": [ 1 ], "names": null }, "frame_index": { "dtype": "int64", "shape": [ 1 ], "names": null }, "episode_index": { "dtype": "int64", "shape": [ 1 ], "names": null }, "index": { "dtype": "int64", "shape": [ 1 ], "names": null }, "task_index": { "dtype": "int64", "shape": [ 1 ], "names": null } }, "total_episodes": 1, "total_frames": 779, "total_tasks": 1, "chunks_size": 1000, "data_files_size_in_mb": 100, "video_files_size_in_mb": 200, "data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet", "video_path": "videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4", "robot_type": "so_follower", "splits": { "train": "0:1" } } ``` ## Citation **BibTeX:** ```bibtex [More Information Needed] ```This dataset is a robotics dataset created using the LeRobot tool, focusing on a multiple box experiment task. It contains 1 episode and 779 frames recorded at 30 fps. Features include action data (e.g., joint positions), observation state data (117 features covering joint positions, current, load, velocity, torque, etc.), image data (from top and corner cameras with 480x640 resolution, 3-channel video), and metadata (e.g., timestamp, frame index, episode index). The robot type is so_follower, with data stored in parquet format and videos in mp4 format. The dataset is suitable for research in robot control, perception, and reinforcement learning.

搜集汇总

数据集介绍

构建方式

该数据集名为multiple_box_experiment_20260527_183717，是在机器人操作领域，通过LeRobot框架采集的演示数据集。数据采集使用so_follower机器人，总计包含1个episode、779帧数据，采样频率为30帧/秒。数据分为训练集（比例为1:1），以parquet和mp4格式存储底层数据，并配备meta/info.json元数据文件，详细描述了特征定义、数据路径及视频编码参数等关键信息。

特点

数据集拥有多模态的传感器融合特征，涵盖六维关节控制指令（action）、包含位置、电流、负载、速度、扭矩及模型估计值等117维观察状态（observation.state），以及来自顶部和角落双视角的RGB视频流（480x640分辨率，AV1编码）。特征空间覆盖关节动力学、动力学模型估计及末端执行器力量信息，为精细的机器人操控学习提供了空前丰富的数据维度。

使用方法

使用者可借助LeRobot库加载该数据集，支持通过HuggingFace上的可视化工具在线浏览或下载后本地处理。数据按视频与表格分块存储，可依据episode_index和frame_index索引帧数据。建议采用随机采样或时间连续切片构建训练批次，配合数据增强与归一化，适用于训练机器人控制策略、学习演示动作或进行仿真到现实的迁移研究。

背景与挑战

背景概述

该数据集由Hugging Face LeRobot团队于2026年5月27日创建，旨在推动机器人操作领域中模仿学习与行为克隆技术的发展。数据集聚焦于一种多箱操作实验，通过SO-100机械臂在真实环境中执行拾取与放置任务，采集了包括关节角度、速度、力矩及视觉信息在内的丰富多模态数据。其核心研究问题在于提供高保真度的机器人演示数据，以支持从状态-动作映射中学习复杂操作技能，从而推动具身智能在工业与日常场景中的自主决策能力。该数据集以Apache-2.0许可公开，采用LeRobot数据格式，为机器人社区提供了一个可复现、易扩展的研究基准，对验证新颖算法和跨平台迁移研究具有重要参考价值。

当前挑战

该数据集面临的挑战涉及多维层面。首要挑战在于领域问题：多箱操作任务要求机器人具备精细的物体辨识、抓取策略与顺序规划能力，而现实环境的动态变化与物体位姿不确定性对模仿学习算法的泛化性能构成严峻考验。其次，数据集构建过程中需克服高精度标定与同步难题——确保多视角视觉流（如顶视与角落相机）与关节传感器数据在30Hz帧率下的严格时序对齐，同时补偿机械臂动力学模型中未建模的摩擦与惯性扰动。此外，单集数据仅含779帧，样本规模受限，可能限制复杂模式的捕捉，而多模态特征的高维性（117维状态向量）也对模型的特征提取与降维能力提出更高要求。

常用场景

经典使用场景

multiple_box_experiment_20260527_183717数据集源自LeRobot框架，专为机器人操作学习而设计。其经典使用场景集中于模仿学习与强化学习的训练与评估，研究者可利用其中包含的高频（30fps）多视角视觉观测（顶部与角落摄像头）和详尽的状态信息（涵盖关节位置、速度、力矩及扭矩估计等117维特征）来构建端到端的机器人控制策略。该数据集尤其适用于研究复杂操作任务，如抓取、堆叠和搬运箱子等精细动作，为验证算法在真实物理环境中的泛化能力提供了标准化的测试平台。

实际应用

在实际应用层面，该数据集可直接用于训练工业机械臂（如so_follower）执行精密装配、分拣和物料搬运任务。企业可借助这些示范数据快速部署自适应机器人系统，减少手动编程成本，并提升在非结构化环境中的作业灵活性与鲁棒性。此外，其高保真的传感器记录（包含电流、扭矩及力传感器数据）也为预测性维护、故障诊断以及人机安全协作机制的设计提供了现实依据，展现出从实验室原型向智能制造、仓储物流等场景转化的潜力。

衍生相关工作

基于该数据集的结构与内容，衍生出一系列重要研究工作。一方面，利用其多模态特征，研究者开发了针对机器人操纵的视觉-语言-动作模型，推动了具身智能体的发展；另一方面，数据集中含有的外部扭矩估计值促进了接触丰富的操作（contact-rich manipulation）算法研究，如阻抗控制和顺应控制策略的改进。此外，该数据集也被用于元学习与跨任务迁移学习的基准测试，探索如何在不同物体排列和场景下复用运动先验，从而催生了若干数据增强和仿真到现实迁移的新方法，深刻影响了仿人机器人和协作机器人的学习范式。