AirSim

It enables researchers and engineers to train and test autonomous driving algorithms, drone navigation systems, and robotics controllers in simulation before deploying to physical hardware reducing the cost, time, and safety risks associated with real-world testing of autonomous systems.

The simulator provides physics-accurate vehicle dynamics, a range of sensor modalities including cameras (RGB, depth, segmentation), LIDAR, IMU, and GPS, and configurable weather and lighting conditions.

AirSim exposes a Python and C++ API for programmatic control of vehicles and sensors, making it compatible with reinforcement learning frameworks (RLlib, Stable Baselines), computer vision pipelines (OpenCV, PyTorch), and robotics middleware (ROS).

Pre-built environments include urban streets, open landscapes, and custom map support through the Unreal Editor.

AirSim was developed by Microsoft Research and is open-source under the MIT license, making it freely available for academic research, education, and commercial development.

It has been widely adopted in autonomous vehicle research programs at universities and automotive companies, and served as a benchmark environment for multiple autonomous driving competitions.

While Microsoft Research archived the project in 2022, the codebase remains actively used and forked by the research community, with community-maintained versions continuing development.