Waymo’s self-driving simulators do 100 years of driving every day

0
400
Waymo is training its self-driving algorithms from home with simulations.

It takes human drivers about 50 hours of behind-the-wheel practice before they can get their driver’s license. While that doesn’t automatically make someone a good driver (it certainly doesn’t) it does give them enough time to learn the basic rules of the road and start making driving judgments.

Artificial intelligence (AI) enabled self-driving programs need significantly more practice. Fortunately, Waymo’s simulators are able to acquire 100 years of driving experience every single day. That training hasn’t stopped because drivers can’t actually be out on the roads. So, yes, even AI is working from home during the coronavirus pandemic.

Driving Experience

Due to the COVID-19 pandemic, Waymo has been forced to temporarily suspend its self-driving test programs across the country. However, that doesn’t mean that it intends to delay its plans for fully autonomous vehicles.

Advertisement
Manage your supply chain from home with Sourcengine

Although on-road testing is an important part of the process, simulations are also valuable training tools for self-driving cars. They allow Waymo’s team to train its AI models far faster than they can with real-world testing.

In just one 24-hour period, Waymo is able to simulate 100 years of driving. To date, the company claims to have simulated 15 billion miles. To put that figure in context, it’s equivalent to making 31,394 round-trip journeys to the moon.

A company representative said in a statement, “Waymo’s simulation technology allows us to create completely simulated scenarios using data from what we’ve seen and experienced driving in the real world.”

The statement goes on to say, “In simulation, we can manipulate the environment around our Waymo Driver to gauge how it will respond to various situations like continuous oncoming traffic while making an unprotected turn.”

That is valuable experience that Waymo’s AI doesn’t always get when it’s training on a real road. Since the simulations can be programmed to replicate any event, the engineers are able to put their virtual driver in harm’s way without worrying about actually causing an accident. Doing that in real-life obviously isn’t feasible.

It is accomplishing this by compiling highlights from its 20 million miles of real on-road test data. By putting the AI through these scenarios over and over again, it gains experience at handling challenging situations.

Making Home Work

It’s extremely impressive that Waymo has been able to keep up this pace of testing despite the fact that workers are conducting their business from home. Alphabet, Waymo’s parent company, ordered all of its employees to work remotely due to COVID-19.

For engineers dealing with resource-intensive simulation tools, working at home could be a nightmare. Fortunately for those on the Waymo team, they work for one of the world’s most advanced technology companies.

Waymo has been able to rapidly develop web-based versions of its most important tools. This has allowed engineers to keep testing and simulating from home without the need for ultra-powerful hardware.

When Waymo is able to get back on the road, it will be very interesting to see how these countless hours of simulation pay off.

LEAVE A REPLY

Please enter your comment!
Please enter your name here