Anthony Levandowski is a pioneer of autonomous vehicle systems. He developed a self-driving motorcycle in 2004, joined Google’s driverless car team early on and left to help start Otto when he decided trucking was a better place for the technology to develop.
Otto was sold to Uber, and the 38-year-old UC Berkeley-trained engineer left under a cloud when Google’s self-driving car company accused Uber of stealing trade secrets. The case was settled with no finding of wrongdoing. Levandowski declines to discuss the matter but said in an interview that he does not believe he’s done anything unethical.
But he does want to talk about autonomous driving, and the reversal in thinking he has had since leaving his former employers. Levandowski now believes that full autonomy with the Lidar-based sensing tech most systems depend on remains a distant goal. But the time is right, he believes, for real-world adoption of a limited system that uses the best of the proven technologies available.
That’s the philosophy behind Pronto AI, Levandowski’s newest company. The name means “ready artificial intelligence,” and Levandowski says Pronto will be ready to start outfitting commercial trucks with its camera-based system within a few months.
Levandowski discussed his concerns about the state of autonomous driving technology and explained Pronto’s system and promise in a recent interview with Trucks.com. Here’s an edited version of that session.
You’ve gone from pushing for full autonomy to being a bit of a skeptic. Why?
I am still guilty of being excited about the future and of hoping to see this technology come out. But I’ve learned a lot in the last several years and realize that the really exciting part is delivering benefits through real products in the short term. But shifting to truck driving on the highway with nobody in it, that’s not going to happen in the next several years.
Yet you just did a coast-to-coast drive hands free with a version of your system.
It was validation for us that the sensors we’re putting on the trucks are capable of doing the entire thing. What’s missing is the software that does this in a reliable, repeatable way.
You started Pronto to capitalize on those years it will take to get to full automation?
Today there’s not even computer-controlled steering on any truck. It’s a very big gap between fixing that problem and having trucks that are fully driverless. We want to offer the best technology in that segment and I think right now we can make the human driver safer. But there’s a lot of steps to get to fully autonomous.
Computer-controlled steering. The product we’re launching now keeps the truck centered in the lane. After that we’ll do things like automatic lane changes. We want to do navigation, where we’ll be able to go from one off-ramp to another off-ramp and take a variety of roads in between. We also think that there is a lot of opportunity in providing fatigue reduction for the drivers.
By driving the truck automatically?
Imagine that a driver is behind the wheel and on duty but hasn’t actually driven because the system is doing that. Those hours wouldn’t be counted against the driver’s legally permitted driving time, and the truck potentially could stay in motion for the full 14 hours of on-duty time, even though the human may have driven manually just five hours. It would reduce fatigue, and maybe even let drivers be home a night more often, because they could make a round trip in a 14-hour window that they couldn’t do before when they could only run the truck for 11 of those hours.
That’s in the future. What’s Pronto offer now?
We’re building a Level 2 self-driving truck kit for fleets and drivers to get the benefits of the technology that exists today rather than waiting for the promise of driverless trucks years from now. We’re focusing on that by making the technology affordable. We are interested in deploying real products that add safety and value right now, and we’ll upgrade over time to do more and more.
A Level 2 system isn’t what people think of as self-driving, is it?
On the highway, you can turn the system on, and it will keep the truck centered in the lane and maintain the proper headway but also brake all the way to a stop due in to traffic jam or if emergency braking is necessary. It still requires the driver to pay attention and monitor the system. But it also does driver monitoring, to confirm that the driver is looking forward, checking mirrors. It looks for cellphones, to make sure they’re not using one. The system’s six cameras also monitor the driver’s arms to make sure they’re not putting their hands behind their head when they need to be near the wheel. We don’t require them to hold the steering wheel, but we do expect them to be responsible.
How do you keep drivers engaged when they don’t have to do anything for hours at a time?
We provide driver training, and the system monitors the drivers and will give them prompts if they’re not checking their mirrors and watching the road. It they don’t correct their actions, the warning keeps going off and the truck will actually slow down and attempt to pull over and get off the road, because we assume at that point that the driver is incapacitated.
Truck makers will adopt autonomy technology as it advances. Will that make Pronto unnecessary?
If all we did was what we’re doing today, I agree with you. But we’ll add value for truck owners if we have the best system out there. And at $5,000, we’re maybe two orders of magnitude less expensive than some of the offerings out there. Manufacturers make their own engines, but customers still buy engines from some other supplier because they like that other engine better. I would love to partner with a truck maker and have Pronto be an option. Let their customers decide. We already have more orders than we can initially supply, and that says something about our value.
Is there tech touted now that you think shouldn’t be part of the discussion?
Daimler said it has stopped working on truck platooning because it doesn’t work. I agree with them on that. That’s an example where a leading manufacturer tried it, it didn’t work out, they’re calling it a day. But I wouldn’t dismiss any new technology that’s still being developed. You’ve got to give them time. I could be wrong and going straight to driverless could be the solution. I just have a hard time imagining us going from trucks that don’t have electric power steering straight to the driverless truck.