ICLR'21 C-Learning: Horizon-Aware Cumulative Accessibility Estimation

Authors: Panteha Naderian, Gabriel Loaiza-Ganem, Harry J. Braviner, Anthony L. Caterini, Jesse C. Cresswell, Tong Li, Animesh Garg [paper]

Environment:

The code was developed and tested on the following python environment:

python 3.7.4
pytorch 1.4.0
numpy 1.17.2
gym 0.15.7
mujoco 1.5

Instructions:

1. In order to train the models run the scripts in the experiment folder:

python experiment/env_name -s 21

The trained models will be saved in runs/env_name directory.

To train the model for other environments you should write similar scripts.

2. For visualizing and saving the simulations:

python evaluate/sim.py FetchPickAndPlace-v1
python evaluate/sim.py HandManipulatePen-v0

3. To plot the learning curves:

python evaluate/training_curve.py env_name

4. For final evaluations on the given metrics:

python evaluate/eval.py env_name

Citation

If you find this code useful in your research, please cite the following paper:

@inproceedings{naderian2020c,
title={C-Learning: Horizon-Aware Cumulative Accessibility Estimation},
author={Naderian, Panteha and Loaiza-Ganem, Gabriel and Braviner, Harry J and Caterini, Anthony L and Cresswell, Jesse C and Li, Tong and Garg, Animesh},
booktitle={International Conference on Learning Representations},
year={2020}
}