BETA
This is a BETA experience. You may opt-out by clicking here

More From Forbes

Edit Story

New Data Platform Helps Autonomous Vehicles Learn Rules Of The Road Directly From Cities

This article is more than 5 years old.

Image by Inrix

Using software and sensors, autonomous vehicles (AVs) have several ways to comprehend the world around them. Cameras detect objects and read road signs, HD maps tell an AV where it is in relation to other objects and artificial intelligence and machine learning help make sense of it all.

But it’s still an imprecise process. “If there’s a highly automated vehicle pulling up to an intersection and its 50% sure there's a stop sign, 25% sure there's a yield sign and 25% sure there's a do not enter sign, that’s basically useless,” says Avery Ash, head of autonomous mobility for Inrix.

The traffic and mobility data company is filling this information gap with its new AV Road Rules platform, a tool that lets cities digitize their traffic rules and restrictions via an open API. Respectively, Inrix AV Road Rules allows companies conducting testing programs to obtain verified street sign and other traffic infrastructure information directly from transportation authorities.

“Cities will be able to go through the process of validating and managing their traffic rules,” Ash adds. “And vehicle operators don’t have to only rely on sensor-based machine learning methods of learning the rules of the road.”

While machine learning and HD mapping for AVs are moving at a rapid pace, they’re still time-consuming, resource-intensive and inexact methods of gathering road data. “HD maps are really good at getting to the centimeter level of understanding that there’s some object at this precise location, and machine learning will try to find out what it is” Ash notes.

“But that machine learning is fed by humans and you have teams of people going through the process of saying, ‘That’s a stop sign and that’s a fire hydrant,’” he adds. “The algorithm gets better and better at recognizing signs, but ultimately it can get close to 100% but never 100%. AV Road Rules is an additional data layer that complements machine learning and HD maps.”

Road Rules not only allows AV operators to get exact locations of traffic signs and other roadway infrastructure from cities, but feeds the information back to cities in a symbiotic relationship. It also fulfills a requirement for AV operators to know local traffic rules and restrictions before they begin testing. “When you’re a vehicle operator and applying for an autonomous vehicle testing permit on public roads, one of the questions is, how do you abide by local traffic rules,” says Ash.

Seven cities and road authorities and four AV operators have signed on to use the AV Road Rules platform. The cities include Austin, Texas, Boston and Cambridge, Massachusetts; Portland, Maine; the Regional Transportation Commission of Southern Nevada, including Las Vegas; and Transport for West Midlands and Transport Scotland in the UK. Automakers and operators planning to use the platform include Jaguar Land Rover, May Mobility, nuTonomy and those running Renovo’s Aware platform.

Ash says Inrix expects to expand to a dozen cities and AV operators before the end of year. “Our initial partners will help us with refining and expanding version two of the platform, and that’s one of the benefits for getting in on the ground floor,” he adds.