Sample-Efficient Reinforcement Learning via Conservative Model-Based Actor-Critic

This is the code of paper Sample-Efficient Reinforcement Learning via Conservative Model-Based Actor-Critic. Zhihai Wang, Jie Wang*, Qi Zhou, Bin Li, Houqiang Li. AAAI 2022. [arXiv]

Requirements

• Python 3.6.9
• PyTorch 1.10
• tqdm
• gym 0.21
• mujoco 1.50

pip install -r requirements.txt

Reproduce the Results

1. For example, run experiments on Ant

python scripts/run.py configs/ant.json

Citation

If you find this code useful, please consider citing the following paper.

@article{wang2021sample,
  title={Sample-Efficient Reinforcement Learning via Conservative Model-Based Actor-Critic},
  author={Wang, Zhihai and Wang, Jie and Zhou, Qi and Li, Bin and Li, Houqiang},
  journal={arXiv preprint arXiv:2112.10504},
  year={2021}
}

Remarks

We will release our data reported in our paper soon.

Other Repositories

If you are interested in our work, you may find the following papers useful.

Model-Based Reinforcement Learning via Estimated Uncertainty and Conservative Policy Optimization. Qi zhou, Houqiang Li, Jie Wang.* AAAI 2020. [paper] [code]