Moham126/eval_tower-of-hanoiv3-adjusted-150-100

--- 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=Moham126/eval_tower-of-hanoiv3-adjusted-150-100"> <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", "robot_type": "so_follower", "total_episodes": 1, "total_frames": 512, "total_tasks": 1, "chunks_size": 1000, "data_files_size_in_mb": 100, "video_files_size_in_mb": 200, "fps": 30, "splits": { "train": "0:1" }, "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", "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_lift.pos", "elbow_flex.pos", "wrist_flex.pos", "wrist_roll.pos", "gripper.pos" ], "shape": [ 6 ] }, "observation.images.front": { "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 } }, "observation.images.topDown": { "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 } }, "observation.images.actionCam": { "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 } }, "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 } } } ``` ## Citation **BibTeX:** ```bibtex [More Information Needed] ```This dataset is a robotics dataset created using the LeRobot tool, specifically designed for the so_follower robot type. It contains 1 complete episode with a total of 512 frames, covering 1 task. The data is stored in parquet format, with a total data file size of 100MB and video file size of 200MB, at a frame rate of 30fps. Features include action data (e.g., shoulder pan, shoulder lift, elbow flex, wrist flex, wrist roll, and gripper positions, represented as 6 float32 values), observation state (joint positions identical to actions), and image observations from three perspectives: front, topDown, and actionCam, each with a resolution of 480x640 pixels and 3-channel color video. Additionally, metadata such as timestamp, frame index, episode index, overall index, and task index are included. The dataset is suitable for research and development in robot control, reinforcement learning, or computer vision tasks.

搜集汇总

数据集介绍

构建方式

该数据集依托LeRobot框架构建，采用SO-100 follower机械臂采集单次任务演示。全程以30帧每秒的速率同步记录六自由度关节位置与三路视觉信号，经分块压缩为Parquet与AV1视频格式，形成包含512帧的标准化机器人学习数据。数据按训练集统一划分，并遵循v3.0版数据规范组织目录结构。

特点

数据集聚焦于汉诺塔任务，以单回合、多视角同步观测为显著特征。除六维关节状态与动作向量外，还提供前视、俯视及动作相机三路视频，帧率与物理时间戳严格对齐。其精简的样本规模适用于算法快速验证与视觉-动作联合建模研究，所有数据均以开放许可发布。

使用方法

研究者可借助LeRobot工具链直接加载数据集，通过解析Parquet文件获取状态与动作序列，并解码对应视频帧用于策略学习或模仿学习。数据支持在Hugging Face可视化空间中交互浏览，亦可用于训练视觉运动策略，使用时需遵循Apache-2.0许可协议并引用来源。

背景与挑战

背景概述

在具身智能与机器人学习领域，高质量真实世界操作数据集的稀缺长期制约着策略泛化能力的提升。eval_tower-of-hanoiv3-adjusted-150-100数据集依托LeRobot框架构建，以SO-Follower机械臂为采集平台，记录了汉诺塔任务在调整参数条件下的单次完整操作轨迹，共512帧、30fps，涵盖前视、俯视及动作相机三路视觉观测与六维关节状态及动作。该数据集虽规模有限，却延续了自2023年以来LeRobot社区推动低成本开源机器人数据共享的路线，为研究精细操作中的多模态表征与模仿学习提供了可复现的评测基准。

当前挑战

该数据集所针对的领域问题在于复杂序列操作任务中的长程依赖与接触密集操作，要求策略在有限示范下理解并复现多层圆盘搬运与放置的时序逻辑，其挑战在于单次演示难以覆盖任务执行中的状态分布多样性，易导致策略过拟合。构建过程中，多相机同步采集与关节控制信号的时间对齐需在低成本硬件上保持稳定，动作空间与视觉观测的精确配准亦受到机械臂重复定位误差和视频编码延迟的干扰，如何在数据规模与采集质量之间取得平衡构成显著难题。

常用场景

经典使用场景

在机器人学习与具身智能领域，该数据集构成了一个典型的基于视觉的模仿学习评估基准。其核心使用场景聚焦于让配备平行夹爪的SO-follower机械臂在模拟或真实环境中执行汉诺塔（Tower of Hanoi）拼装任务。数据集通过前置、俯视和动作相机三个视角，以30帧每秒的速率同步采集了机械臂的关节位置、末端执行器动作及多模态视觉观测，为训练端到端的视觉-动作映射策略提供了精细的时序对齐样本。

解决学术问题

该数据集直面机器人操作中长时序任务规划与多步推理的学术难题。汉诺塔问题天然涉及递归逻辑与状态依赖性，要求智能体不仅识别当前盘片布局，还需预测后续移动序列。通过提供包含执行动作与观测状态的完整轨迹，它为研究模仿学习中的误差累积、视觉泛化及任务分解提供了受控实验平台，推动了从简单抓取到复杂逻辑操作的研究范式演进。

衍生相关工作

基于该数据集，研究者已衍生出多项围绕视觉模仿学习与任务规划的经典工作。例如，结合LeRobot框架开发的策略网络在汉诺塔任务上验证了扩散策略与Transformer架构的效能；部分工作进一步利用其多视角特性探索了三维场景理解与动作分块方法。此外，该数据集还被用于评估离线强化学习算法在稀疏奖励环境下的表现，催生了针对长时序任务的分层策略研究。