CARLA

An open-source, Unreal Engine-based self-driving simulator commonly used for closed-loop testing of driving policies.

CARLA was released by Alexey Dosovitskiy, German Ros, Vladlen Koltun, and colleagues at the first CoRL in 2017. It provides open-source digital assets for city road layouts, buildings, and vehicles, and lets users flexibly configure cameras, lidar, depth, and GPS sensors, along with weather, lighting, and other traffic participants. The original paper used it to compare three driving approaches: a traditional modular pipeline, an end-to-end imitation-learning model, and an end-to-end reinforcement-learning model. It has since become a standard platform for closed-loop self-driving evaluation, with an accompanying autonomous-driving Leaderboard. According to its website, version 0.10.0 (December 2024) upgraded to Unreal Engine 5.5, while the Unreal Engine 4-based 0.9.x line is still maintained, with 0.9.16 released in September 2025; the code is MIT-licensed and the assets are CC-BY licensed.

ExampleThe CARLA autonomous-driving Leaderboard has competing driving systems drive in closed loop along fixed routes and traffic scenarios, automatically scoring them on route completion and rule violations.

Also called

CARLA Simulator

Related

Autonomous Driving · Simulator · Game Engine · Closed-Loop Evaluation · End-to-End · LiDAR

Sources

CARLA: An Open Urban Driving Simulator (arXiv 1711.03938)
CARLA Simulator 官网 (Chinese)
carla-simulator/carla (GitHub)

As of

2026-09