AirSim: An Advanced Simulation Platform for Autonomous Vehicles
AirSim is a remarkable simulation platform that has made significant contributions to the field of AI research and experimentation. It was initially created by Microsoft Research in 2017 and has since served as a common means to share research code and test new ideas in aerial AI development and simulation.
The platform is built on Unreal Engine and now also has an experimental Unity release. It supports a wide range of vehicles, including drones and cars, and offers various features such as software-in-the-loop and hardware-in-loop simulations with popular flight controllers. AirSim also exposes APIs for retrieving data and controlling vehicles in a platform-independent manner, enabling users to experiment with deep learning, computer vision, and reinforcement learning algorithms.
One of the key advantages of AirSim is its flexibility. Users can generate training data in multiple ways, either by simply pressing the record button or by accessing the APIs for more customized data logging. Additionally, the platform offers features like manual control using a remote control or arrow keys, as well as programmatic control through APIs. It also includes a Computer Vision mode for collecting images and exploring the scene without the need for vehicles or physics.
With the advancements in technology and the growing needs of the aerospace industry, Microsoft is set to release a new simulation platform, Project AirSim, in the coming year. While the original AirSim will be archived, users will still have access to its code. Project AirSim is expected to provide an end-to-end platform for safely developing and testing aerial autonomy through simulation, offering advanced capabilities such as safety, code review, testing, and AI functionality.
In conclusion, AirSim has been a valuable tool for the AI research community, and its successor, Project AirSim, holds great promise for the future of aerial autonomy and simulation.
