CARLA is used for the development, training and validation of autonomous driving systems. Like AirSim, Carla is developed with Unreal Engine, using a server and multi-client architecture. In terms of scenarios, CARLA provides open-source digital resources for creating scenarios for autonomous driving (including urban layouts, buildings, and vehicles) and several scenarios built from these resources for autonomous driving test training. At the same time, CARLA can also use VectorZero’s road construction software RoadRunner to create scenes and supporting high-precision maps, and also provides a simple map editor. CARLA can also support flexible configuration of sensors and environment, it supports multi-camera, lidar, GPS and other sensors, and it can also adjust the light and weather of the environment. CARLA provides a simple automatic behavior simulation of vehicles and pedestrians, and also provides a complete set of Python interfaces, which can control vehicles, signal lights, etc. in the scene, which is used to facilitate co-simulation with the automatic driving system to complete the decision-making system. and end-to-end reinforcement learning training.