A Comparative Review of Autonomous Driving Simulation Platforms: Carla, AirSim, and Gazebo

Engineering notes

This study provides a comparative review of three primary open-source platforms: Carla, AirSim, and Gazebo.

Chinese explanation / 中文解读

中文解读待补充：本站会优先为端到端自动驾驶、BEV感知、3D目标检测、轨迹预测、路径规划、LiDAR感知等高价值论文补充中文说明。

Original abstract

The rapid evolution of autonomous driving has made high-fidelity simulation essential for validating safety-critical algorithms. Since real-world testing is cost-prohibitive and riskyespecially for edge casesdevelopers rely heavily on virtual environments. This study provides a comparative review of three primary open-source platforms: Carla, AirSim, and Gazebo. Instead of a general overview, we evaluate these tools based on architecture, visual fidelity, physics accuracy, and ecosystem support. The analysis highlights clear trade-offs: Carla is the most effective for urban traffic scenarios due to its rich asset library; AirSim excels in perception tasks by leveraging high-end game engines for photorealism; and Gazebo remains the standard for robotics control due to its precise physics capabilities. The findings offer a practical selection guide, emphasizing that the choice of simulator must depend on the specific subsystem under test, such as end-to-end learning versus vehicle dynamics. Finally, the paper discusses current computational limitations and identifies digital twins as the key future trend for bridging the gap between simulation and reality.

Links and sources

• Official / arXiv page

Comments