Tesla Autopilot
Radar readings and camera views are fed into the Autopilot system on Tesla cars. (Credit: Tesla)

Tesla car owners are getting an upgrade to their Autopilot semi-autonomous driving software, but this won’t be your standard software upgrade: CEO Elon Musk says Autopilot 8.0 will put more emphasis on radar readings as well as crowdsourced, networked information about potential hazards on the roadway.

Musk said the upgrade might have prevented the kind of collision that led to the death of a Tesla Model S driver in May. “These things cannot be said with absolute certainty, but we believe it is very likely that, yes, it would have,” Musk told reporters during a teleconference on Sunday.

That accident involved a crash between the Model S and a freight truck that was making a left turn from the opposing direction. Preliminary data suggest that Autopilot’s camera system did not recognized the reflective signature of the truck against the background of a brightly lit sky. As a result, the car smashed with full force into the truck, killing Tesla driver Joshua Brown.

The National Highway Traffic Safety Administration says it’s investigating the accident.

Since then, Tesla has improved the software that processes a separate set of readings from the radar system on Tesla cars. Previously, the Autopilot system used radar readings only to confirm what the cameras were seeing. With the upgrade, radar will become the primary scanning method.

The software update, which will be pushed out wirelessly to Tesla cars over the next couple of weeks, also draws upon database of road data shared by different Tesla vehicles driving over the same road. Without that database, the radar system might have a hard time distinguishing between a harmless but radar-reflective object – for example, an overhead road sign – and a genuine road hazard.

“The big problem in using radar to stop the car is avoiding false alarms,” Tesla explained in a blog posting written primarily by Musk.

To help address that challenge, Tesla says its cars will use “fleet learning”:

“Initially, the vehicle fleet will take no action except to note the position of road signs, bridges and other stationary objects, mapping the world according to radar. The car computer will then silently compare when it would have braked to the driver action and upload that to the Tesla database. If several cars drive safely past a given radar object, whether Autopilot is turned on or off, then that object is added to the geocoded whitelist.”

Over time, Tesla cars will rely increasingly on the database to assess radar blips.

“The net effect of this, combined with the fact that radar sees through most visual obscuration, is that the car should almost always hit the brakes correctly even if a UFO were to land on the freeway in zero-visibility conditions,” Tesla said.

Tesla emphasizes that Autopilot doesn’t yet qualify as an autonomous self-driving system. To underscore that fact, the system will sound a warning when a driver lets go of the steering wheel while traveling at highway speeds. (That Autosteer feature is more forgiving at lower speeds.) If a driver ignores three warnings over the course of an hour, Autosteer will be disengaged until the driver stops the car, shuts it off and restarts it.

Musk said that Tesla let the NHSTA know about the upgrade, and that safety officials “appear pretty happy with these changes.”

In the wake of the fatality in May, Consumer Reports called on Tesla to disable Autosteer until it’s programmed to require drivers to keep their hands on the wheel, and to stop referring to its control system as “Autopilot.” But Musk signaled that Tesla would stick with the system, even though it’s still in beta, and would keep improving it.

“It’s not about going from bad to good,” Musk told reporters. “Things are already good. I think it’s about going from good to great.”

Like what you're reading? Subscribe to GeekWire's free newsletters to catch every headline

Job Listings on GeekWork

Find more jobs on GeekWork. Employers, post a job here.