One of the challenges currently faced by automated driving technology is how to train AI (artificial intelligence) algorithms to make decisions and drive cars. This requires data collection from millions or even billions of miles of driving. andThis is exactlyMicrosoftThe garage project Project Road Runner wanted to solve.
According to information published in the Microsoft Garage Official Blog, Project Road Runner is a research project in the Microsoft Garage in Redmond, Washington. The project is mainly to help the autopilot industry, developers, and researchers to collect data to train AI through simulated reality. As the name suggests, it is close to a realistic road scene in a controlled simulation environment, which makes the test more flexible and minimizes the impact on the actual road.
The above picture from Microsoft's official blog
Since the project has no funds, the team was operating on a free simulator for remotely controlled cars and the Internet at the very beginning. However, they later turned to AirSim, an open-source simulation platform developed by Microsoft Research based on the Unreal engine, which is mainly used for research on drones. AirSim is also an ideal platform for training autopilot algorithms, enabling more accurate data capture in real-life situations.