Dataset for Paper "The EMPATHIC Framework for Task Learning from Implicit Human Feedback"

This is the raw data set for the paper "The EMPATHIC Framework for Task Learning from Implicit Human Feedback", published at CoRL 2020. (project website)

The data set contains human facial reactions recorded while the subjects watching autonomous agents completing a task that determines their reward. This data set contains data recorded in two domains: Robotaxi and RoboticSortingTask (details of each domain can be found in the paper).

This data set file is organized as:

• RobotaxiData/:
  • subject_id/
    • episode_N/
      • [subject_id]_episode_N_[time]_[time_step]_[frame_id].jpg
  • RLstatistics_episode_N.csv: time_step, reward, optimality
• RoboticSortingTaskData/
  • subject_id/
    • [subject_id]_[trajectory_order]_[trajectory_id]
  • trajectory_returns.csv: trajectory_id, final_return

If you find this data set useful for your research, please consider citing our paper:

@inproceedings{cui2020empathic,
  title={The EMPATHIC Framework for Task Learning from Implicit Human Feedback},
  author={Cui, Yuchen and Zhang, Qiping and Allievi, Alessandro and Stone, Peter and Niekum, Scott and Knox, W Bradley},
  booktitle={Conference on Robot Learning},
  year={2020},
  organization={PMLR}
}