I already wrote two posts about the recent automotive day at the Computer History Museum, Silicon Valley Reinvents the Wheel and The California DMV on Autonomy. Today it is the turn of trucks. Given that a truck costs about $200-250K, it is much easier to justify the cost of, say, $25K of electronics than it is in a $40K car.

Autonomous Trucks

There was a sort of Q&A on autonomous trucks, with Jenny Eflsberg of Volvo, and Chuck Price of TuSimple. Stuff in [brackets] are my comments.

Chuck had brought along a video to explain what TuSimple is up to. I think it was assumed that the audience knew that Volvo is a major manufacturer of trucks (and cars, of course). And that they are Swedish.

The video started by pointing out that despite everything you might hear, we currently have a shortage of 50,000 truck drivers. Many existing drivers are also nearing retirement so this problem is likely to get worse. The shortage is predicted to be 125,000 by 2024.

Q: Tell us about the video. I know you are running trucks in Arizona in revenue service.

Chuck: Yes, we have 40 trucks in Tucson. We did a run from Phoenix to Dallas with USPS. We have about 18 shippers that we are moving freight for. I can't name most of them due to NDAs. But USPS [the postal service] and UPS [the FedEx competitor] let us go public.

Q: Do you have safety drivers?

Chuck: Yes. These trucks are fully level 4 with both a safety driver and a safety engineer at all times.

Q: So what's Volvo's approach?

Jenny: Very different from TuSimple. We have the Vera truck, as if you tried to redesign the truck without space for anyone to go with the truck. It is also electric. We are deploying first in Gothenberg. Electric traction means it is not long-haul.

Q: Is it geographically limited or just by the range of the vehicle?

Jenny: It is. We are obsessed with safety and so have a control tower approach. We really want the Vera machine to be independent of the control tower and we have other layers of controls.

[This video was not shown at the meeting, but here is Volvo's Vera]

Q: Back to you Chuck. How quickly do you see this playing out? How do you scale up?

Chuck: We are using revenue runs to validate everything, and we have to operate on the shippers' schedules. We expect to have our validation complete for our first driver out operations in 2021. We expect the first factory-built vehicles in 2023 available for sale.

Q: You'll be working with OEMs to add to an existing vehicle?

Chuck: Yes, we are working with several OEMs. We are not building our own trucks.

Q: Are you a technology company? Or a freight company?

Chuck: Technology. We are working with the OEMs and tier-1s to provide autonomous capability. We are then in the business of operating the fleets purchased by the customers when they are on public roads. We provide other services as needed: we intervene if the vehicle is in trouble. The vehicle is level 4 so it must rely on its own safety system.

Q: Safety drivers?

Chuck: We hope by 2021 to be able to do without safety drivers. It's critical long-term.

Q: Jenny, talk about how important connected vehicles are to Volvo? How do you use that part of the business?

Jenny: The old model was you sell the truck, and everyone knows where to find diesel. But now connectivity is needed since nobody is ready to completely let go with no knowledge of what is going on.

Q: Volvo has technology to monitor all the trucks?

Jenny: Yes, that all feeds into a database. It is super interesting to us to see whether things are correct or not. We are going to more and more intelligence so access to that data is crucial.

Q: What is the basic system TuSimple uses?

Chuck: We are somewhat different but much the same. We are camera-forward: the primary sensor is camera. We do also use lidar and radar. We can go out to about 1000m, which is about twice the distance a commercial truck driver is trained to look.

Q: Driving depot to depot in the rain. Are you using maps?

Cuck: Yes, it is map-based technology. We only drive where we have mapped, but we can operate in heavy rain and other situations where you might assume cameras would be ineffective. Take a look at the videos on the TuSimple channel on YouTube. There is a 1.5-hour ride in the rain compressed into six minutes but otherwise it is not edited.

[Here it is]

Q: Jenny, you are in charge of Volvo's Silicon Valley lab. What are you looking at?

Jenny: Everything. I'm part of connected solutions so we are looking at the digital space and new business models. Other teams are coming over and looking into electrification and autonomy.

Q: Final question, Chuck. On your website you talk about the importance of Amazon AWS.

Chuck: We work closely with Amazon. Our system is highly AI-based and requires significant GPU power for training and other tasks. When we don't have sufficient capacity in-house we scale to AWS. they also hold a lot of data. We have onsite a petabyte of local storage, but push the rest to AWS.

Q: One of your competitors announced they don't want to do any on-road testing.

A: We think on-road testing should be approached very carefully. Safety first. That's why we currently place two people in the vehicle. Other forms of testing, such as simulation and test tracks should be done before the highway. Some tests we would only ever do on the tracks. But we operate on our customers' route and customers' schedules so we can't cherry-pick and make ourselves look good.

Q: Jenny, Chuck, thank you very much.

Sign up for Sunday Brunch, the weekly Breakfast Bytes email.