Microsoft AI simulator expands to autonomous car research

November 22, 2017 // By Rich Pell
Microsoft (Redmond, WA) have added car simulation to their AirSim system for testing the safety of artificial intelligence systems.

The company's AirSim is an open source AI simulator originally available for testing the safety of autonomous drones. Developed as a plug-in for the game development tool suite Unreal Engine , it offers physically and visually realistic environment simulations for research into how AI-based autonomous vehicles can operate safely in the open world.

In addition to several new enhancements for airborne vehicles, AirSim now includes car simulations, new environments, APIs to ease programming, and ready-to-run scripts. By providing an open, community-driven platform for testing algorithms for self-driving cars, the company hopes to help advance the research and development of self-driving vehicles.

The system features a detailed 3D urban environment that includes a variety of diverse conditions, including traffic lights, parks, lakes, and construction sites. Users can test their systems in several types of neighborhoods, including downtown, semi-urban, vegetation, and industrial environments, and on more than 12 kilometers of drivable roads spanning more than 20 city blocks.

Because the car simulation is decoupled from the environment it runs in, users can create an environment for their specific needs. AirSim extensibility also allows researchers and developers to incorporate new sensors, vehicles, or even use different physics engines, says Microsoft.

AirSim's APIs can be used in a wide variety of languages, allowing it to be used with various machine learning tool chains. In addition, says Microsoft, it sees significant opportunities with running multiple instances on AirSim with Microsoft Azure to scale up the training for data-hungry machine learning algorithms.

AirSim has also been made available as a compiled binary release, allowing users to download and start calling its Python APIs to control a vehicle in just minutes. Plans for future releases include new sensors, better vehicle physics, weather modeling, and more detailed realistic environments.

Microsoft/AirSim

Related articles:
AI wars heat up with new Microsoft research lab
Intel, Mobileye: Formula proves safety of autonomous vehicles
'Slaughterbots' video depicts autonomous killer drones
Microsoft Project Brainwave


Vous êtes certain ?

Si vous désactivez les cookies, vous ne pouvez plus naviguer sur le site.

Vous allez être rediriger vers Google.