What has your research revealed as the best use of this technology?
Autonomous trucks can only drive over the long-haul portion of a trip—in this case, highways—because urban environments are just too complex for computers to navigate. So the truck drives from the outskirts of Philadelphia, say, to the outskirts of Chicago. A human driver then meets up with the truck, gets the trailer, and makes the final delivery. In other words, we’re going to need a lot of local drivers.
Some estimates have projected self-driving trucks could cost the industry anywhere from 1.5 to 3 million jobs. Is this what you’ve found?
I think it’s more like a couple hundred thousand jobs lost, but there will also a lot of new jobs created. There has been this idea that the job of anyone in this category of ‘truck driver’ is at risk. But most truck drivers do a good deal more aside from driving: loading and unloading, paperwork, stocking shelves. It doesn’t necessarily make sense to automate the driving function if you’re going to have a whole bunch of labor on either side of that task.
Can you describe these new jobs?
They’re what are called last-mile jobs, and they’re being driven by the speed of e-commerce. The name comes from how products get moved in the final mile of shipping before they arrive with the customer. We should not worry only about job loss; we should also be worried about job quality. The long-haul jobs, as challenging as they are, are still the premium jobs in the industry in terms of what drivers can earn.
What impact might driverless trucks have on the environment?
It really comes down to a matter of policy. Here’s how: I had this idea that we should create urban truck ports. Off peak, long-haul drivers would bring fully loaded trucks back and forth between them, and then local drivers would complete the jobs. With segmentation of the work into local and long-haul, however, there is tremendous potential but a lot of risk. This long-haul self-driving technology could drop total costs by around one-third for highway miles. If we then use energy-efficient technologies in conjunction, we could reduce costs by 40 or 45 percent for highway travel overall, but that means a whole bunch of cargo that now goes by train would start to go by truck. That’s bad for energy consumption. Trains are at least four times as efficient as trucks, plus more trucks will damage highways and muck up rush hour, so, if we don’t put in place policy alternatives that control how much comes off of rail, we could have a real energy nightmare on our hands.
What kind of policy do you mean?
To start, a carbon tax and ambitious fuel-efficiency standards that really promote innovation. And for labor, congestion, and infrastructure, a tax on autonomous driving mileage. Vehicle mileage taxes are a popular idea and are a way to get around the trade-off with fuel economy and funding. In other words, as vehicles become more efficient, fuel tax revenue goes down, which means less funding to maintain infrastructure. Such a tax could make up for some of the potential damage being done to the roads by additional trucks and keep our roads from being clogged by lots of cheap but inefficiently used trucks. It could also provide funds to ease the transition for workers displaced by the new technology.
Are these trucks safe?
These are 80,000-pound vehicles, so safety is absolutely critical. We have to look at the difference between rural and urban miles. Urban miles are really complex. People do all kinds of things on roadways, and humans pick up on those right away. For the computer algorithm to learn those nuances will take a long time, but on the highway miles, it’s a lot simpler. Humans get fatigued, they stop paying attention, and at high speeds that’s really problematic. But computers are never going to fall asleep. Right now, the big challenge is weather because it affects truck performance by interfering with the sensors. It takes trucks about 300 feet to stop in good conditions—that’s about as far as the sensors can see right now—so the sensors have to get better. And then there’s the machine-learning side of it. There are a bunch of areas that all need to move forward before this can work well and predicting that is impossible other than to say that this is not a side project.
You’re a sociologist, but here you’re making economic predications about a transportation sector. How do you mesh those?
It’s been challenging. As a sociologist, I look at what’s happened in the past and what is happening currently. This is about projecting into the future, which most academics are not comfortable doing, yet economists and transportation planners do it frequently. It was a bit of a hurdle for me to get past initially, but I realized I could apply my usual methods to create scenarios based on different assumptions, different possible futures that policy could help us bring about.
How long before we truly see driverless trucks on the road?
There are some debates among the developers, but for the most part they know what they need to do and how they want to do it. You could see self-driving trucks on the highway in some areas, like the U.S. Southwest, within the next five years. But it could be 20 years before we see significant market penetration to the point where you’re going to have declining job opportunities or fewer truckers. And that’s still not local driving and delivery jobs.