Repolode
Submit
metadriverse

metadriverse/metadrive

A lightweight Python driving simulator for reinforcement learning and autonomous driving research, with sensor simulation and Gymnasium-style environments.

View on GitHubWebsite
Stars
1,274
Forks
210
Open issues
103
Last commit
23 days ago

What is it?

MetaDrive is a driving simulator built for AI and autonomy research. It can generate an effectively unlimited number of scenes from different road maps and traffic settings, or load real-world driving logs, which suits research on generalizable reinforcement learning.

The simulator is designed to be easy to install and run on Linux, Windows, and macOS. It provides sensor simulation including point clouds, RGB, depth and semantic images, a top-down semantic map, and first-person views. Environments follow the Farama Gymnasium interface, and the project includes multi-agent scenarios such as roundabouts, intersections, toll gates and parking lots, plus an option to replay nuScenes or Waymo scenarios with reactive traffic.

Key features

  • 01Generates diverse scenes from procedural road maps and traffic settings.
  • 02Loads real-world driving logs from nuScenes and Waymo datasets.
  • 03Provides point cloud, RGB, depth, semantic and top-down map observations.
  • 04Exposes Gymnasium-compatible RL environments.
  • 05Includes multi-agent scenarios such as roundabout, intersection and parking lot.
  • 06Installs with pip and runs on Linux, Windows and macOS.

Use cases

  • Training and evaluating RL agents for generalization across unseen driving scenes.
  • Prototyping perception or planning pipelines with simulated sensor data.
  • Replaying real-world logs from nuScenes or Waymo with reactive traffic for testing.
  • Running multi-agent driving experiments at intersections or roundabouts.

Quick start

  1. 1

    Clone the repository

    git clone https://github.com/metadriverse/metadrive.git
  2. 2

    Enter the project folder

    cd metadrive
  3. 3

    Install MetaDrive in editable mode

    pip install -e .
  4. 4

    Verify the installation by running the test script

    python -m metadrive.examples.profile_metadrive

Things to know

  • Some Python environment setup and a Panda3D-based renderer are needed to run the simulator with graphics.

In collections

Open-source AI agent frameworks & appsReinforcement learning frameworks & training environmentsRobotics & embodied AI projects

Topics

#autonomous-driving#reinforcement-learning#simulation#gymnasium#robotics#multi-agent#python

FAQ

What is metadrive?
A lightweight Python driving simulator for reinforcement learning and autonomous driving research, with sensor simulation and Gymnasium-style environments. MetaDrive is a driving simulator built for AI and autonomy research. It can generate an effectively unlimited number of scenes from different road maps and traffic settings, or load real-world driving logs, which suits research on generalizable reinforcement learning.
Is metadrive free to use? What license does it have?
metadrive is open source under the Apache-2.0 license. Check the license terms before using it in commercial products.
Is metadrive still maintained?
Actively maintained. The last commit to metadrive was on Oct 16, 2026.
What can I use metadrive for?
Training and evaluating RL agents for generalization across unseen driving scenes. Prototyping perception or planning pipelines with simulated sensor data. Replaying real-world logs from nuScenes or Waymo with reactive traffic for testing. Running multi-agent driving experiments at intersections or roundabouts.
How do I install metadrive?
Clone the repository: git clone https://github.com/metadriverse/metadrive.git Enter the project folder: cd metadrive Install MetaDrive in editable mode: pip install -e . Verify the installation by running the test script: python -m metadrive.examples.profile_metadrive
Where can I get metadrive?
The source code and installation instructions are on GitHub: https://github.com/metadriverse/metadrive

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects