carla-simulator/carla

Open-source simulator for autonomous driving research.

What is CARLA?

CARLA is an open‑source simulator built on Unreal Engine that lets researchers develop, train, and evaluate autonomous‑driving systems. It provides a realistic urban environment (roads, buildings, traffic) together with a flexible sensor suite (cameras, LiDAR, radar, GNSS, etc.) and lets you control weather, traffic density, and many other conditions.

Key Features

  • High‑fidelity 3D world powered by Unreal Engine 5.8 (development branch ue58-dev).
  • Ready‑made digital assets – city maps, vehicle models, pedestrians, traffic signs.
  • Customizable sensor suites – attach any combination of cameras, depth sensors, LiDAR, GNSS, IMU, etc., to simulated vehicles.
  • Python API for easy scripting of scenarios, data collection, and integration with ML pipelines.
  • Blueprint library for visual scripting inside Unreal.
  • Ecosystem tools such as the autonomous‑driving leaderboard, Scenario Runner, ROS bridge, and reinforcement‑learning examples.
  • Cross‑platform builds for Ubuntu 22.04/24.04 and Windows 11, with detailed build scripts.

Typical Use Cases

  • Training perception and planning models on synthetic data before fine‑tuning on real‑world recordings.
  • Benchmarking autonomous‑driving stacks via the CARLA leaderboard or custom scenario runners.
  • Testing sensor fusion and control algorithms under varied weather, lighting, and traffic conditions.
  • Research on reinforcement learning, imitation learning, and simulation‑to‑real transfer using the provided RL and COIL‑train repositories.
  • Integration with ROS 2 or other robotics middleware through the ROS‑bridge.

Getting Started

  1. Clone the UE5.8 branch:
    git clone -b ue58-dev https://github.com/carla-simulator/carla.git CarlaUE5
  2. Run the setup script (CarlaSetup.sh on Linux, CarlaSetup.bat on Windows) – it installs prerequisites, downloads the Unreal Engine source (you need a linked Epic‑Games GitHub account), and builds the simulator.
  3. Build the project with CMake/Ninja (see the README for exact commands).
  4. Launch the editor or run headless simulations via the Python API.

Documentation & Community

  • Full docs: https://carla-ue5.readthedocs.io
  • Build guides, first‑step tutorials, API reference, and Blueprint library are all linked from the README.
  • Community support via GitHub Discussions, Discord, and a dedicated forum.

License

  • Core CARLA code: MIT License.
  • Assets: Creative Commons BY.
  • Dependencies follow their own open‑source licenses (Apache‑2.0, MPL‑2, etc.).

If you use CARLA in a publication, cite the original CoRL 2017 paper.

Related

  • Project
    carla-simulator/scenario_runner

    A traffic scenario definition and execution engine for the CARLA simulator used to test and prepare autonomous driving agents.

  • Project
    ucla-mobility/OpenCDA

    An open-source co-simulation framework for Cooperative Driving Automation (CDA) that integrates CARLA and SUMO to research multi-vehicle cooperation and V2X communication.

  • Project
    louiszengCN/CarlaAir

    A unified simulation infrastructure that integrates CARLA and AirSim into a single process to enable high-fidelity, synchronized air-ground embodied intelligence research.

  • Project
    Thinklab-SJTU/Bench2Drive

    A closed-loop dataset and benchmark for autonomous driving systems using the CARLA simulator to evaluate driving scores, success rates, and specific driving abilities.

  • Project
    autonomousvision/carla_garage

    A complete open-source starter kit for the CARLA leaderboard 2.0, providing datasets, training code, and pre-trained TransFuser++ models for end-to-end autonomous driving.