CARLA Simulator: An Open-Source Revolution in Autonomous Driving
CARLA is an open-source simulator for autonomous driving research, developed to support the development, training, and validation of autonomous driving systems. It offers open-source code, protocols, and digital assets (urban layouts, buildings, vehicles) for free use. The platform allows flexible sensor suite specification, environmental condition control, and full control over static and dynamic actors. It also features map generation and much more.
Highlighted Features
- Scalability: A server-multi-client architecture enables multiple clients (on the same or different nodes) to control different actors simultaneously.
- Flexible API: A powerful API provides comprehensive control over simulation aspects, including traffic generation, pedestrian behavior, weather conditions, sensors, and more.
- Autonomous Driving Sensor Suite: Configure diverse sensor suites, including LIDARs, multiple cameras, depth sensors, and GPS.
- Fast Simulation: A rendering-disabled mode offers fast execution of traffic simulation and road behaviors when graphics aren't needed.
- Maps Generation: Easily create custom maps using the ASAM OpenDRIVE standard and tools like RoadRunner.
- Traffic Scenarios Simulation: The ScenarioRunner engine allows defining and executing various traffic situations based on modular behaviors.
- ROS Integration: Seamless integration with ROS via a ROS-bridge.
- Autonomous Driving Baselines: Pre-built autonomous driving agents (including AutoWare and a Conditional Imitation Learning agent) are available.
Core Features
- Quickstart: Easy installation and setup guides are available.
- Actors: Interact with various entities within the simulation, such as vehicles, pedestrians, and traffic signals.
- Sensors: A wide array of realistic sensor models (cameras, LIDAR, RADAR) are included, along with access to ground truth data (semantic segmentation and depth information).
- Traffic Manager: Control non-player characters (NPCs) to challenge your autonomous driving agent.
- ROS Bridge: Enables seamless connection with the Robot Operating System.
Tutorials
- First Steps in CARLA: A beginner-friendly tutorial to get started.
- Bounding Boxes: Learn how to access ground truth bounding boxes for vehicles and map features.
- Creating Custom Maps: Create your own maps using CARLA's asset library or by importing your own assets.
- Creating Custom Vehicles: Design and integrate your own custom vehicles.
- Recording Simulations: Record and replay simulations to compare results across different sensors and configurations.
Latest News
- CARLA CVPR Autonomous Grand Challenge winners announced! (June 10, 2024)
- CARLA teams up with Neya Systems to upgrade to Unreal Engine 5 (March 26, 2024)
- NVIDIA Omniverse Cloud APIs (March 18, 2024)
- OASIS Sim from Synkrotron (December 12, 2023)
- CARLA 0.9.15 Release (November 10, 2023)
Get Involved
Support the project by starring the GitHub repository!