Amazon-Backed Aurora Readies An Air-Traffic Control System For Self-Driving Cars

Self-driving cars aren’t quite ready for large-scale commercial deployment, but as the technology advances most companies developing it plan multiple layers of safety. Aurora, a startup that counts Amazon as an investor, thinks a system of remote monitors acting much like air-traffic controllers is one way to help ensure public acceptance of its robotic chauffeurs.

The Palo Alto-based company, founded by three autonomous tech veterans from Google, Tesla and Uber, is investing in a system it’s calling “teleassist” as it works to perfect the software, computers, sensors and vision system that allow its test fleets in the Bay Area and Pittsburgh to handle a range of road conditions. The plan is for trained technicians at remote facilities to access a vehicle’s sensors when needed and offer suggestions and guidance for unusual developments. Aurora vehicles would safely pull over to the side of the road when making such calls for advice.

While companies such as Phantom Auto and autonomous truck startup Starsky Robotics are testing remote control systems, Aurora’s teleassist doesn’t work that way, says cofounder and chief product officer Sterling Anderson.

“Our system will alert remote teleassist personnel when the need arises, either because a rider or another user of the vehicle in a logistics network has requested it or because the vehicle has called for help—‘I see something that falls outside my comfort zone. Tell me at a high level what you think I should do,’” Anderson tells Forbes. Even so, the advice provided by remote assistance personnel “cannot be part of the functional safety for the vehicle,” he says.

He declined to provide specific details for how much Aurora is investing in the monitoring system, when its teleassist facilities would open or when it would go live.

Rather than allowing direct control of our vehicles, Aurora is building a teleassist API that can pass information between teleassist specialists and its self-driving software system.

Led by former Google Self-Driving Car Project leader Chris Urmson, Aurora has raised about $700 million since emerging from stealth mode in January 2018. Amazon’s connection formed in February 2019 when the retail behemoth joined Aurora’s Series B round, which raised more than $530 million. Aurora has begun testing autonomous delivery vans and trucks, in addition to passenger vehicles, though Anderson declined to discuss whether an Amazon self-driving partnership is in the works.

Teleoperation has long been used to steer space probes and drones, but hasn’t been fully adapted for robotic cars. Waymo, the commercial self-driving company set up by Google in 2016, also uses human monitors at its facilities in suburban Phoenix to keep tabs on and assist its robotic ride service minivans but who don’t remotely take control of vehicles. Although Tesla, which Elon Musk says will be an early leader in self-driving tech, collects large amounts of on-road data from drivers using its Autopilot system, the company doesn’t appear to use remote guidance personnel.

Anderson, who helped design Tesla’s Autopilot system, wouldn’t discuss his former employer’s approach.

The latency of cellular network signals and “situational awareness” difficulties for remote personnel, understanding all the critical circumstances a vehicle is contending with, create too many challenges to safely take over driving control, according to Anderson. Still, an approach like teleassist may help get self-driving fleets into operation sooner as the onboard software continues to get better.

“Our approach to self-driving is safely, quickly, broadly. And if we want to get in market and make a difference sooner, we can do that safely if we introduce a system like this,” Anderson says. “Chances are our system will behave just fine in certain situations, but until and unless we’ve proven that it can do so safely with a strong statistical argument, we’re going to take the conservative path.”

Form:forbes