dataset_info:
features:
- name: map_name
dtype: string
- name: trajectory_id
dtype: int64
- name: steps_data
dtype: string
- name: num_steps
dtype: int64
splits:
- name: train
num_bytes: 307264902
num_examples: 624
- name: evaluation
num_bytes: 46719495
num_examples: 96
- name: all_data
num_bytes: 353984397
num_examples: 720
download_size: 148078966
dataset_size: 707968794
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
- split: evaluation
path: data/evaluation-*
- split: all_data
path: data/all_data-*
task_categories:
- robotics
language:
- en
tags:
- human-robot-interaction
- multi-agent
- reinforcement-learning
- collaboration
Moving Out: Physically-grounded Human-AI Collaboration Dataset
This repository contains human-human interaction data collected for the Moving Out: Physically-grounded Human-AI Collaboration benchmark.
Moving Out introduces a new human-AI collaboration benchmark designed to evaluate embodied agents' ability to adapt to physical actions and constraints in an environment. The dataset includes human-human interaction data collected for two tasks within this environment, covering various collaboration modes such as moving heavy items together and coordinating actions around corners. It is crucial for understanding how models adapt to diverse human behaviors and unseen physical attributes in physically-grounded scenarios.
Project Page: https://live-robotics-uva.github.io/movingout_ai/
Code Repository: https://github.com/live-robotics-uva/Moving-Out
Sample Usage
You can visualize trajectories from this dataset using the provided dataset_to_video.py script available in the project's GitHub repository.
Save trajectory as MP4:
python dataset_to_video.py -f ShuaKang/movingout_task2 -m HandOff -t 4 -v video
Show trajectory in popup window:
python dataset_to_video.py -f ShuaKang/movingout_task2 -m HandOff -t 4 -v human
Use -m for map name and -t for trajectory ID. Remember to replace ShuaKang/movingout_task2 with the specific dataset ID you are using (e.g., ShuaKang/movingout_task1 or ShuaKang/movingout_task2).
Citation
If you use this dataset in your research, please cite the following paper:
@article{li2025moving,
title={Moving Out: Physically-grounded Human-AI Collaboration},
author={Li, Shua and Liu, Zi and Guo, Ruohan and Bai, Min and Li, Erran and Levine, Sergey and Lin, Dahua},
journal={arXiv preprint arXiv:2507.18623},
year={2025}
}