Article Image


Microsoft's AI Simulator Allows for Autonomous Car Research

by Justin Tejada
Testing autonomous vehicles is an enormous undertaking. Designing software that can pilot a vehicle -- a massive task unto itself -- is only the beginning.

Building a physical automobile with all the necessary hardware components for autonomy is another huge hurdle.

After that comes the challenge of finding a closed course to test on and then acquiring a permit to test on public roads. Of course, established companies that have spent years and millions of dollars developing autonomous vehicles have already cleared all these hurdles. And that's how it should be. No one wants their nextdoor neighbor testing a DIY AV around the neighborhood cul-de-sac. But the barrier to entry shouldn't be so high as to discourage all newcomers.

Open-source autonomous vehicle software platforms mean that developers do not need to reinvent the wheel every single time they want to build something new and also encourage contributions from all corners of the industry.

Earlier this year, Microsoft open-sourced AirSim, a platform for testing the artificial intelligence systems of autonomous vehicles. At launch, it was used exclusively to test aerial drone systems. Now, Microsoft has announced that AirSim will start allowing users the opportunity to test and refine driverless car systems.

"We aim to make [the] various aspects of developing self-driving cars available to a broader group of researchers by providing an open, community-driven platform for testing those algorithms," researchers Ashish Kapoor and Shital Shah wrote in a post announcing the development. "The new version of AirSim includes car simulations, new environments, APIs to ease programming and ready-to-run scripts to jump start your research."

The detailed 3D urban environment in AirSim has more than seven miles of drivable roads that cover the equivalent of 20 city blocks. The landscape includes things like traffic lights and construction sites, so developers can replicate real-world conditions that autonomous vehicles are likely to encounter. The entire environment can be customized to accommodate different needs and future updates may include factors such as varying weather conditions.

For autonomous vehicle developers, virtual testing of systems can be just as important as real-world miles for improving their software. The Atlantic recently reported on how Waymo puts the brains of its driverless cars through intense learning processes without driving a single physical mile.

"In that virtual space, they can unhitch from the limits of real life and create thousands of variations of any single scenario, and then run a digital car through all of them," Alexis Madrigal wrote. "As the driving software improves, it's downloaded back into the physical cars, which can drive more and harder miles, and the loop begins again."

The faster that loop gets recycled, whether it's on a Microsoft platform or a Google one or some other yet-to-be-developed software, the faster autonomous vehicles will be ready to take the road at scale.


Automated Driving: How Government Can Help

Governments at all levels have key roles to play in the convergence of the transportation, technology, and infrastructure that will be necessary to enable automated driving. Jeff Stewart, AT&T Assistant Vice President for Public Policy, will discuss several key interrelated policy initiatives: smart cities, small cell deployments, FirstNet for first responders, broadband deployment, and V2X technologies. He will also share how policies can help protect against security risks and help ensure the safety of drivers, passengers and pedestrians.

Like us on Facebook