This repository contains the data of all 27 subjects of the user study of the paper: "On the importance of environments in Human-Robot Coordination". Matthew Fontaine*, Ya-Chuan Hsu*, Yulun Zhang*, Bryon Tjanaka and Stefanos Nikolaidis. RSS 2021.
You can view the data and replay the collaborated human-robot game play here.
The data of the user study are under the user_study/results directory. There are 27 directories in total, each containing the data of one subject. The data of each subject is in its corresponding <subject_ID>/human_log_refined.csv file.
It is useful to set up a conda environment with Python 3.7 using Anaconda:
conda create -n overcooked_lsi_user_study python=3.7
conda activate overcooked_lsi_user_study
To complete the installation after cloning the repo, run the following commands:
cd overcooked_lsi_user_study
pip install -e .
Use the following command to run the replay:
python replay_user_study.py -l <subject_ID> -type <lvl_type>
<subject_ID> is the directory name (note: not the full path) of the subject and <lvl_type> is the type of the corresponding level that you want to replay. <lvl_type> must be one of the following:
even_workloads-0
even_workloads-1
even_workloads-2
uneven_workloads-0
uneven_workloads-1
uneven_workloads-2
high_team_fluency-0
high_team_fluency-1
high_team_fluency-2
low_team_fluency-0
low_team_fluency-1
low_team_fluency-2
For example, if you want to replay the trace of the subject 1 playing the level even_workloads-0, use the following command:
python replay_user_study.py -l 1 -type even_workloads-0
The overcooked_ai_py directory is adopted from this project by the
Center for Human-Compatible AI.