To facilitate the research of generalizable reinforcement learning, we develop an open-source, highly efficient and flexible driving simulator MetaDrive, which holds the following key features:

• Compositional: It supports generating infinite scenes with various road maps and traffic settings for the research of generalizable RL.
• Lightweight: It is easy to install and run. It can run up to 300 FPS on a standard PC.
• Realistic: Accurate physics simulation and multiple sensory input including Lidar, RGB images, top-down semantic map and first-person view images. The real traffic data replay is also supported.

We construct a variety of RL tasks and baselines in both single-agent and multi-agent settings, including benchmarking generalizability across unseen scenes, safe exploration, and learning multi-agent traffic.

@misc{li2021metadrive,
      title={MetaDrive: Composing Diverse Driving Scenarios for Generalizable Reinforcement Learning},
      author={Quanyi Li and Zhenghao Peng and Zhenghai Xue and Qihang Zhang and Bolei Zhou},
      year={2021},
      eprint={2109.12674},
      archivePrefix={arXiv},
      primaryClass={cs.LG}
}