ergoCub Humanoid Puzzle Manipulation: Teleoperation Logs of Unsolving and Solving a Puzzle (IIT AMI)

Provider
Artificial and Mechanical Intelligence (AMI) lab, Istituto Italiano di Tecnologia (curated by Giovanni Fregonese)
License
CC-BY-4.0 (source)
Size
24 files
Format
Parquet (in sampled files)
Stored at
Hugging Face (https://huggingface.co/datasets/ami-iit/puzzle_manipulation_datasets)
Access
Free and public. No account with the data provider is needed.
Category
Robotics
Source
https://www.science.org/doi/10.1126/scirobotics.adh3834

About this dataset

Logged teleoperation data from ergoCub, the humanoid robot of the Istituto Italiano di Tecnologia, manipulating a puzzle: in one experiment the robot first takes the puzzle apart (unsolve01) and then solves it (solve01). Data were collected by the AMI research line with the immersive avatar teleoperation stack developed for the ANA Avatar XPRIZE competition, described in the iCub3 avatar system paper (Science Robotics, 2023). Each folder holds Parquet tables of synchronized robot signals, including desired and measured joint states from the walking and manipulation controllers and timestamped RealSense RGB and depth image paths. 22 Parquet files, about 7.7 GB (measured 2026-11-21); read with pandas, pyarrow or Hugging Face datasets. The dataset was made to develop autonomous manipulation algorithms for ergoCub, such as imitation learning of bimanual humanoid manipulation from teleoperated demonstrations. It contains no personal data. Released under CC BY 4.0 by IIT; cite the iCub3 avatar system paper, Science Robotics, doi:10.1126/scirobotics.adh3834.

How to load this dataset in Python

from datasets import load_dataset
ds = load_dataset("ami-iit/puzzle_manipulation_datasets")
print(ds)

Data structure

File types in sampled files: Parquet

Files

Example file: solve01/train-00000-of-00012.parquet

Analyze this dataset without downloading it

On Scigantic, ergoCub Humanoid Puzzle Manipulation: Teleoperation Logs of Unsolving and Solving a Puzzle (IIT AMI) opens in a hosted JupyterLab notebook with the data already mounted, so you can read files where they are stored instead of copying them first. Each notebook includes an AI assistant that can see the mounted files and write and run analysis code. Open the dataset page to launch a notebook, or browse the full catalog.

This page is also available as Markdown and as JSON.

Related Robotics datasets