This code accompanies the paper "Plan To Predict: Learning an Uncertainty-Foreseeing Model for Model-Based Reinforcement Learning".
- Install MuJoCo 1.50 at
~/.mujoco/mjpro150and copy your license key to~/.mujoco/mjkey.txt - Clone
P2P
git clone https://github.com/ZifanWu/Plan-to-Predict.git
- Create a conda environment and install Plan-to-Predict
cd src/Plan-to-Predict
conda env create -f environment/gpu-env.yml
conda activate p2p
pip install -e .
python src/main_p2p.py --num_epoch 150
The optimal parameters are contained in .src/configs/ folder.
@article{wu2022plan,
title={Plan To Predict: Learning an Uncertainty-Foreseeing Model For Model-Based Reinforcement Learning},
author={Wu, Zifan and Yu, Chao and Chen, Chen and Hao, Jianye and Zhuo, Hankz Hankui},
journal={Advances in Neural Information Processing Systems},
volume={35},
pages={15849--15861},
year={2022}
}